Skip to main content

Documentation

User manual

All 57 guides in reading order. Use your browser's Print to save it as a PDF; every heading links back to its own page.

Back to the index

1. Get started

Accounts, the first project and how the platform thinks.

1.1 Your home page

What needs you today: projects, tasks with deadlines, meetings, recent manuscripts, notifications, learning and teams — in one screen.

What is on it

  • Your projects — the six you touched most recently, with open tasks, manuscripts and members.
  • My tasks — everything assigned to you in any project, overdue first, then by deadline; badges count overdue and due-this-week items.
  • Recent manuscripts and upcoming meetings (next seven days) across your projects.
  • Notifications — the five latest and the unread count.
  • Continue learning, your teams and recently in your library.

Everything loads in one request and refreshes every minute while the page is open.

Getting started

Until you have created a project, added references, started a manuscript, worked with someone and completed your profile, a checklist at the top links to each step. It disappears when all five are done.

Home is also where sign-in lands. Press G then P to jump to Projects, or use the command palette (⌘K) for anything else.

1.2 Getting started

Create an account, finish the four-step onboarding, start a project and connect your own AI agent to it.

Create your account

Sign up with your e-mail address and a password of at least ten characters, or with a passkey. Passwords are checked against a list of common choices before they are accepted, so pick a phrase of several words.

A verification link goes to your inbox. You can use the platform straight away; the banner at the top disappears once the address is confirmed. Verification unlocks project invitations by e-mail, digests and account recovery.

Institutions with single sign-on can pick “Continue with your institution” and enter a work e-mail address; the administrator has to register the identity provider first.

The four onboarding steps

  1. Who you are: display name, position, institution, department and country. This is how colleagues see you on projects and in messages.
  2. What you work on: research interests, methods, discipline and career stage. Discovery and journal recommendations use these as defaults.
  3. ORCID: type your iD (the check digit is validated) or sign in with ORCID. Linking lets you import publications and keeps citation metrics current.
  4. Publications: paste DOIs or PubMed IDs, or import everything from ORCID in one click.

Every step can be skipped and finished later under Settings. “Skip for now” takes you straight to your projects.

Start a project

  1. Open Projects and choose New project.
  2. Give it a name, a field and a type (standard research, systematic review, meta-analysis, clinical trial…). The type decides which tabs are switched on.
  3. Invite collaborators from the Team tab with a link or by e-mail; titles such as principal investigator, co-investigator or research assistant map to permissions.

A project holds everything for one piece of work: team, tasks, meetings, files, the reference library, evidence, screening, analyses, manuscripts, submissions and the conversations your AI agent has about it.

Connect your AI

Hikma runs no models of its own: the AI you work with is the one you already use. Open the project's Your AI tab, pick your client — Claude Code, Claude Desktop, claude.ai, ChatGPT, Codex, Cursor, VS Code, Windsurf, Zed or Cline — and copy the command it shows you. Signing in through the client (OAuth) or a personal key both take one step, and the panel confirms as soon as your agent makes its first call.

Then ask in your own client, in plain language: “in project X, find randomised trials of amiodarone prophylaxis after cardiac surgery since 2015, add the most relevant ones and draft the evidence section”. Your agent searches the backbone sources through Hikma, adds the works to the project, reads the full texts the platform can fetch and writes into your manuscript as tracked changes. Everything it does appears back in the Your AI tab as a conversation.

Nothing your agent writes is applied silently, and it cannot invent a citation: citations are objects bound to works in the project library, and numbers are bound to the evidence they came from. Read Your AI agent next.

Phone, tablet and desktop

The web app adapts to narrow screens: side panels become drawers, the manuscript workspace keeps a floating action button for its panels and comments, and the bottom bar gives one-tap access to home, discover, projects, notes and your profile.

Desktop applications for Windows, macOS and Linux and a mobile app are on the roadmap; they share the same account and data.

1.3 How the platform thinks

Evidence first, sources over summaries, tracked changes instead of overwrites, and one place for a project.

Evidence before prose

Claims are built from anchored evidence: a quotation, the page it comes from and the work it belongs to. Evidence is extracted first and prose written second — by you, or by the AI agent you connect — so every sentence in a draft can be traced back.

When a source cannot be read (paywalled, no open-access copy), the platform says so rather than guessing from an abstract.

Sources are canonical

A paper is one record however many times it is found. OpenAlex, Crossref, PubMed, Europe PMC, arXiv, Semantic Scholar, ClinicalTrials.gov, Unpaywall and OpenCitations are queried through one registry and de-duplicated into a single work with all its identifiers.

Nothing is overwritten

Manuscript edits by a co-author or by a connected AI agent are tracked changes with an author and a time; accepting one keeps the proposer's name on the text. Versions can be compared and restored. Deleted text never counts as a match in find and replace, and citations survive edits because they are objects, not typed text.

One place per project

Tasks, meetings, files, the reference library, evidence tables, screening decisions, analyses, manuscripts and submissions live in one project with one team and one permission model. You never export from one tool to import into another.

Your data stays yours

By default the platform sends nothing to a language model: your own agent does the reasoning on your own subscription, and Hikma records what it did in the project. Where an administrator has switched the platform's own AI on, the provider keys are held by the administrator, never by you, and every model call is logged under your account with its cost. Profiles have visibility settings, messages have contact preferences, and you can export or delete your account from Settings.

1.5 Every page of the platform

A page-by-page map of Hikma Research: what each screen is for and which guide explains it in depth.

Top level

PageWhat it is forGuide
HomeYour dashboard: projects, tasks due, recent manuscripts, notifications, learningHome dashboard
ProjectsEvery project you belong to; create, filter, archiveProjects
My tasksEverything assigned to you across projects with due datesMy tasks
LibraryYour personal reading library with notes, tags and PDFsLibrary
NotesPersonal and project notes with templates and foldersNotes
MessagesDirect and group conversations, requests, attachmentsMessages
NotificationsOne inbox with filters, mutes and digest settingsNotifications
SearchPeople, papers, projects, manuscripts, notes, teams and courses at onceSearch
DiscoverResearchers, publications across sources, projects, teams, claim your publicationsHome feed and Discover
FeedPosts from people you follow and activity in your projectsHome feed and Discover
TeamsGroups with channels, forum, files and linked projectsTeams
IdeasIdea board: pitch, apply, turn into a projectIdea board
MentorshipFind a mentor, ask, shared goals and notesMentorship
ConferencesDirectory, deadlines, abstract submissionsConferences
PDF requestsAsk authors for full texts and answer requestsPDF requests
AcademyCourses, resources, quizzes, certificates, blogAcademy
SettingsProfile, research, publications, CV, notifications, security, integrations, memory, activity, subscriptionSettings
ProfileYour public page: overview, publications, impact, network, portfolioPublic profile
Documentation and SupportGuides, the one-page manual, the support formGetting help and giving feedback

Inside a project

Under the hub bar one line says what the open tool is for, with a “Learn how” link straight into this manual; every systematic-review stage carries the same line. Confirmations and short questions (delete, rename, a snapshot label) open inside the page, never as browser pop-ups.

TabWhat it is forGuide
OverviewProgress, deadlines, workload, quick linksProjects
Tasks · Calendar · MilestonesBoard, timeline, assignments, approvals, phasesTasks and milestones
Meetings · Canvas · DiscussionScheduling with notes and decisions; whiteboard; threads and pollsMeetings and canvas
TeamMembers, titles (permissions), invitationsTeam and roles
Files · ActivityFolders and uploads; who did whatFiles
SourcesSearch, plan, snowball, acquire full textSources and search
RequestsPapers Hikma could not obtain and is asking you for, with every identifier and what was triedSources and search
ReferencesWorks in the project with notes and tagsLibrary
EvidenceAnchored extractions and verificationEvidence
ReviewProtocol, strategies, screening, full text, PRISMA, risk of bias, GRADESystematic reviews
AnalysisMeta-analysis workbench, statistics, sample size, randomisationMeta-analysis · Statistics
Data · Graphs · FormsDatasets, figure generator, data-collection formsDatasets · Graphs · Forms
ManuscriptsThe editor: pages, ribbon, citations, tracked changes, authorship, checks, exportThe manuscript editor
Reporting checklistCONSORT, STROBE, PRISMA and 32 more, attested item by itemReporting checklists
Journal finderWhere to submit, ranked from your manuscriptJournal finder
Peer reviewInternal review rounds with scores and a decisionInternal peer review
SubmissionsTargets, pre-flight, agreements, package, reviewer responsesSubmissions
Your AIConnect Claude Code, Codex, Cursor or any MCP client, and follow what it does hereYour AI agent
AI auditThe platform's own runs and model calls; disclosure — only where platform AI is onAI audit and disclosure
SettingsProject record, visibility, danger zoneProjects

Administration (administrators only)

Overview, Users, Analytics, Audit log, System, AI & keys, AI costs, Quality, Evals, Flags, Billing, Settings, Mentorship, Reports, Support — see the Administration guide.

1.6 Keyboard shortcuts

Move around the app and the manuscript editor without touching the mouse.

Everywhere

Ctrl/CmdKCommand palette: jump to a project, page or document
Ctrl/Cmd/Show this list
GHHome
GPProjects
GLLibrary
GNNotes
GMMessages
GTTeams
EscClose dialogs, drawers and menus

Manuscript editor

Ctrl/CmdB / I / UBold, italic, underline
Ctrl/CmdShiftXStrikethrough
Ctrl/CmdAlt1…3Heading levels
Ctrl/CmdShift7 / 8Numbered or bulleted list
Ctrl/CmdShiftCInsert a citation
Ctrl/CmdShiftMInsert a comment
Ctrl/CmdFFind and replace (regular expressions supported)
Ctrl/CmdShiftEnterPage break
/Slash menu: headings, lists, tables, figures, equations, page break (AI actions appear only where platform AI is on)
Ctrl/CmdZ / ShiftZUndo / redo (collaborative)
Ctrl/CmdSSave a named version

Tables

Tab / ShiftTabNext / previous cell (adds a row at the end)
Enter in the last rowNew row
Right-click or ⋮Row and column tools, borders, shading, sort

Touch devices

Long-press a word to select it; the formatting bubble appears above the selection. Swipe from the right edge in a manuscript to open the side-panel drawer — changes, comments, authors and your AI.

1.7 Frequently asked questions

Short answers about accounts, data, collaboration, AI and troubleshooting.

Accounts and access

  • Can I use ORCID to sign in? Yes, once it is linked to your account (Settings → Research or Security). ORCID does not share your e-mail, so the first sign-up is with e-mail or a passkey.
  • I lost my authenticator. Use one of the backup codes shown when you enabled two-factor. If those are gone too, contact support from a verified e-mail address.
  • Can one account belong to several institutions? Yes. Institution is a profile field; projects and teams are independent of it.

Data and security

  • Where are my files? In object storage under your project, served only to project members with a valid session.
  • Who can read my manuscripts? Members of the project, according to their role. Administrators see metadata for support and billing, not your text.
  • Which model reads my documents? By default none of ours: your own agent reads them over MCP, under its provider's terms and on your own subscription. Where an administrator has switched the platform's own AI on, it is the provider chosen under Admin → AI & keys, and each run shows the model and its cost.
  • Can I delete everything? Settings → Security → Delete account removes your profile, memberships, messages and keys. Shared project content stays with the project, attributed to a former member.

Collaboration

  • Why did a message become a request? People you share no project with can only send a request first; the recipient's contact preference decides who may write at all.
  • Two of us edited the same paragraph. The editor is collaborative in real time; both edits are kept as tracked changes with their authors.
  • Can I mute a busy project? Yes: project menu → Mute notifications, or the bell on any notification.

AI

  • Where is the AI? In your own client. Connect Claude Code, Codex, Cursor, Claude Desktop, ChatGPT or any MCP client from a project's Your AI tab; it reaches your projects through Hikma's MCP server and does the thinking on your own subscription.
  • Does it cost me anything on Hikma? No. Your provider bills your agent's reasoning as it does today; everything the agent does through Hikma — searching, fetching full texts, extracting evidence, writing, screening, pooling, analysing — is free.
  • Can it write the discussion for me? It proposes paragraphs as tracked changes, each sentence citing works in your library and each number bound to its evidence. You decide what to accept, and the accepted text keeps its authorship.
  • Does it remember me? Only what you confirm under Settings → Assistant memory. An agent's inferred memories are proposals until you keep them.
  • Why is there no Assistant tab, Workflows or AI writing button? Those run on the platform's own models and appear only where an administrator has switched platform AI on. Then each run has a budget, and when it is reached the assistant summarises what it did and proposes the next step.

Troubleshooting

  • A page looks stale. Live updates arrive over one event stream per tab; if your network dropped, reload once.
  • A DOI could not be resolved. Check it on doi.org; preprints and books are sometimes missing from the backbone sources. Add the work manually from the References tab.
  • Exports fail. Very large manuscripts with many images can exceed the export limit; split figures into a supplement or lower image resolution.

1.8 Getting help and giving feedback

Where to find answers, how to report a problem so it can be fixed quickly, and how feature requests are handled.

Finding answers

  • Search the documentation (⌘K anywhere, or the search box on the Documentation page); the one-page manual is printable.
  • Every page of the app has a Help link that opens the matching guide.
  • The FAQ covers accounts, exports, budgets and how AI works here.

Reporting a problem

  • Use the Support form (signed in or not): say which project and page you were on, what you expected and what happened; attach a screenshot if you can. Requests are tracked and answered by a person; you see the thread under Your requests.
  • For anything about a specific project, the project's Discussion tab reaches the whole team.
  • Security concerns go to the support address marked confidential; they are handled first.

Roadmap and platforms

  • The web application is the reference; desktop applications for Windows, macOS and Linux and a mobile app share the same account and data and arrive as they are ready.
  • Institutional sign-in (SSO), shared budgets and private catalogues are available to organisations — contact us through the Support form.
  • Feature requests are welcome through the same form; the most requested ones are announced in the home feed when they ship.

2. Project workspace

Everything a project holds: team, tasks, files, meetings, library.

2.1 My tasks and reminders

Tasks assigned to you across every project, grouped by project, with inline status changes and automatic deadline reminders.

The list

Sidebar → Tasks. Filters: Open, Overdue, Due this week, In review, Done; narrow to one project when you work in several. Change a status from the dropdown without opening the task; the arrow opens it inside its project with comments, subtasks, attachments and time.

A task counts as yours when you are the primary assignee or one of the co-assignees. Below the list, Deadlines in your projects shows tasks due in the next fourteen days in projects you belong to, whoever they are assigned to.

Reminders

  • When a task of yours falls due within 24 hours you receive one notification.
  • When it becomes overdue you receive one more.
  • Both respect your notification preferences (in app and e-mail) and project mutes; changing the due date resets them.

Assigning a task notifies the assignee immediately; completion, review requests, approvals and comments notify the people involved.

2.2 Projects

A project is the unit of work: one team, one permission model, and every tab a study needs.

Creating a project

  1. Projects → New project.
  2. Name it, pick the field and the type. Types switch tabs on: a systematic review shows Review and PRISMA, a meta-analysis shows Analysis, a clinical trial shows Data.
  3. Optionally set a target date and a short description; both appear on the overview.

You become the principal investigator. Invite others from the Team tab.

The tabs

TabWhat it holds
OverviewProgress, deadlines, workload and quick links
TasksBoard, timeline, assignments, approvals and time tracking
CalendarDue dates and milestones by day and month
MeetingsSchedule, RSVP, notes, decisions and action items
MilestonesPhases with linked tasks
FilesFolders and uploads
CanvasWhiteboard with templates and medical icons
DiscussionThreads, polls and decisions with @mentions
TeamMembers, titles and invitations
ActivityEverything that happened, by whom
SourcesSearch, plan, snowball and acquire full text
ReferencesWorks in this project with notes, tags and folders
EvidenceAnchored extractions and verification
ReviewScreening, PRISMA, risk of bias, GRADE
AnalysisMeta-analysis and plots
DataDatasets and sandboxed analyses
ManuscriptsWrite with tracked changes and object citations
SubmissionsJournal targets, reviewer responses, pre-flight and packages
Your AIConnect your own agent over MCP and follow what it does here
SettingsProject record, visibility, danger zone

Tabs you never use can be hidden per project under Settings → Tabs; the command palette (Ctrl/Cmd + K) still reaches them.

Visibility and archiving

Projects are private to their members. Archiving hides a project from lists while keeping everything readable; restoring is one click. Deleting is permanent and asks the principal investigator to type the project name.

The project menu also mutes notifications from that project without leaving it.

Activity and notifications

Every change is recorded on the Activity tab with its author. Members are notified of mentions, assignments, meetings, review decisions and the proposals an AI agent leaves in a manuscript, according to their own notification settings.

The Overview

Every project opens on its Overview, and every card reads live project data: quick actions (new task, upload, dataset, manuscript, meeting, invite, connect your AI), weighted progress, task counts with alerts for unassigned, blocked and overdue work, resource counts (manuscripts, references, evidence, files, datasets, form responses, shared links, members), the review pipeline with PRISMA counts for review projects, completion velocity over the last 30 days with a projected finish, team workload, deadlines, the next meeting, milestones, the latest analysis outputs, links shared by the team, and recent discussion and activity.

Finding your way in a project

The project bar has two rows: sections on top — Plan (Overview, Tasks, Calendar, Meetings, Milestones), Collaborate (Files, Canvas, Discussion, Team, Activity), Research (Sources, References, Evidence, Review, Analysis, Data, Forms), Write & publish (Manuscripts, Peer review, Reporting checklist, Journal finder, Submissions) and AI (Your AI; plus Workflows and AI audit where an administrator has switched the platform's own AI on) — and the tools of the open section below. The section holding the page you are on opens by itself; Settings sits at the right end.

2.3 Team and roles

Invite collaborators, give them titles that map to permissions, and manage what each person can do.

Inviting people

  1. Team tab → Create invite link, or invite by e-mail address.
  2. Choose the title the invitee gets on joining. Links can expire and be revoked.
  3. People without an account sign up first and land in the project after onboarding.

Leaders are notified when someone joins. Pending e-mail invitations show under the members list until accepted.

Titles and what they allow

TitleTypical rights
Principal investigatorEverything, including deleting the project
Co-investigatorEdit everything, manage members, cannot delete the project
Statistician / methodologistEdit analyses, data, evidence and manuscripts
Research assistantAdd and edit content; cannot change membership
ReviewerComment and screen; cannot edit manuscripts
ObserverRead only

Titles are the single source of permissions across every tab and every MCP client: an agent connected by a reviewer cannot edit a manuscript either.

Changing roles and leaving

Leaders can change a title or remove a member; the person is notified. Anyone can leave a project from the project menu; the last principal investigator must hand over first.

2.4 Tasks, milestones and calendar

Plan the work: boards and timelines, assignments with approvals, subtasks, comments, time tracking and milestones.

Tasks

  • Board, list and timeline views; drag between columns to change status.
  • Assignees, due dates, priorities, labels and weights (weights drive the project progress ring).
  • Subtasks with their own checkboxes; comments with @mentions; attachments from the project files.
  • Volunteer for or claim open tasks; leaders approve completed work or request a revision.
  • Timers record time per task; totals appear on the overview workload chart.

Milestones

Milestones group tasks into phases with a target date. Progress is computed from linked tasks; overdue milestones surface on the overview and in the calendar.

Calendar

Due dates, milestones and meetings by day and month. Click a day to add a task or meeting; drag to reschedule.

Who is told

Assignment, approval, revision requests, comments and near deadlines notify the people involved, following their own notification settings.

2.5 Meetings, canvas and discussion

Schedule meetings with RSVP, notes and action items; sketch on a shared canvas; decide things in threads and polls.

Meetings

  • Schedule with attendees, location or link and an agenda; attendees RSVP and get reminders.
  • Live notes during the meeting; decisions and action items become tasks with one click.
  • Recurring meetings and time-zone-aware times for distributed teams.

Canvas

A collaborative whiteboard per project with templates (study flow, PICO map, timeline) and a medical icon set. Export as PNG or SVG for slides.

Discussion

Threads by category (question, decision, announcement, idea) with replies, reactions, polls and @mentions. Mark a thread as decided to pin the outcome; decisions are searchable and appear in the activity log.

2.6 Files

Project files in folders with previews, versions and the same permissions as everything else.

Uploading

Drag files onto the Files tab or use Upload. PDFs, Office documents, spreadsheets, images and data files up to 200 MB are accepted. Files are content-addressed, so re-uploading the same document does not duplicate it.

Each person has a storage allowance — 5 GB by default, set by the administrator — and everything you upload anywhere on the platform counts against yours; the Files header and Settings → Activity show how much is left, and deleting a file frees it again.

PDFs of papers belong in References (they are parsed, cited and quoted); use Files for protocols, ethics approvals, forms and figures.

Folders and search

Create folders, move files between them and search by name. Every file shows who uploaded it and when; the Activity tab records downloads of sensitive files.

Datasets

Tabular data uploaded on the Data tab is profiled and scanned for identifiers before anyone can analyse it; see the Data page.

Datasets, manuscripts and outputs as files

The root of Files also lists the project's datasets (with rows and variables), manuscripts and saved analysis outputs, and the search box finds them. They open in their own editors — Data, the manuscript editor, the Output viewer — while uploads keep their folders, moves and renames.

2.7 Datasets: the data editor

An SPSS-style dataset editor inside every project: typed variables, a case grid, computed variables, Select Cases, pivots, descriptives, import and export.

Creating and importing

  • Project → Data → New dataset: name it and define the first variables (label and type); ten empty rows are created so you can start typing.
  • Import file: CSV, TSV or Excel are read in your browser. Types are detected from the values (number, integer, text, date, yes/no); check names, labels, types and measures, skip columns you do not need, then import. Values that do not fit a type become missing and are counted before you confirm.
  • Every dataset has a change history: who added rows, changed variables, computed or filtered, and when.

Data view

  • Double-click a cell, press Enter or start typing to edit; Tab and Enter move on; Delete clears; Ctrl+C copies the selected block as TSV and Ctrl+V pastes a block from a spreadsheet.
  • Every value is validated against its variable type before it is saved; other members' saves appear within seconds.
  • Click a row number to select rows (Shift for a range) and delete them; add rows from the toolbar; drag a column edge to resize; switch on Value labels to see labels instead of codes.

Undo and redo (Ctrl+Z / Ctrl+Y) work across cell edits; structural changes are listed in the History panel.

Variable view

One row per variable: name (snake_case, used in formulas), label, type, measure (scale, ordinal, nominal, identifier), role, decimals, value labels (1 = Female) and user-defined missing values (999, ranges). Changing a type converts the stored values; anything that cannot be converted becomes missing and is reported. Reorder with the arrows; deleting a variable removes its values from every case.

Computed variables, Select Cases, pivots and descriptives

  • Compute: a formula over other variables — arithmetic, comparisons, AND/OR/NOT, IF, MISSING, ROUND, MEAN, DATEDIFF, AGE and more — previewed on the first rows and kept up to date when the inputs change.
  • Select cases: a condition such as age >= 18 AND sex = "F"; excluded cases are greyed out and left out of descriptives, pivots and exports until the filter is cleared.
  • Pivot: rows by one variable, optional columns by another, count or a summary (mean, median, SD, min, max, sum) of a numeric variable; copy as TSV.
  • Descriptives: n, mean, SD, SE, 95% CI, median, quartiles, range, skewness and kurtosis for numeric variables; frequencies with percentages for categorical ones.
  • Export CSV honours the active filter and, optionally, writes value labels instead of codes.

2.8 Library and references

Your reading collection and each project's reference list: import, de-duplicate, tag, annotate, cite and export.

Personal library and project references

Your Library is private: papers you collect across projects. A project's References tab lists the works that belong to that project. Adding a paper to a project from your library links the same canonical record; nothing is copied.

Adding papers

  • Import RIS, BibTeX, NBIB, EndNote XML or CSV; duplicates are merged by DOI, PMID and title.
  • Paste DOIs or PubMed IDs; metadata is fetched from the backbone sources.
  • Add manually when a record has no identifier (grey literature, reports).
  • Search on the Sources tab and add results to the project directly.

Tags, collections and smart collections

Tag freely; group into collections; smart collections update themselves from a saved filter (for example “tag = pending AND year ≥ 2020”). Notes attach to a work and travel with it into every project.

Reading and annotating

Open a PDF in the reader to highlight, comment and extract evidence anchored to the page. Annotations belong to you or to the project depending on where you opened the paper.

Export

Export any selection as RIS, BibTeX or CSV, or a formatted bibliography in Vancouver, AMA, APA, NEJM, Lancet, BMJ, Nature or Chicago.

Retractions and corrections

Hikma re-asks Crossref, OpenAlex and PubMed about your papers on a schedule (daily by default) and records four kinds of notice separately: retracted, superseded by another version, an expression of concern, and a published correction. Each one is stored with the source that said so, the DOI of the notice, and the time it was read — so a work with no badge is one that was checked, not one nobody looked at. The project's Sources tab lists both, including how many works have never been checked.

A notice propagates. Evidence units taken from that work become superseded: they no longer bind a number in any manuscript, the claims resting on them are marked, and cached readings of the paper drop out of the packets the assistant writes from. A bound number whose source was retracted or superseded is the error NUM_RETRACTED_SOURCE; a concern or a correction is the warning NUM_CONCERN_SOURCE; a citation to such a work is CIT_RETRACTED or CIT_CONCERN. Any release candidate built on the work goes stale and must be rebuilt.

Files you have already exported and packages you have already sent are never changed or resubmitted. Instead the affected version is recorded and named — the candidate, the filename and its sha256, the journal and the manuscript number — together with the action you have to take: contact the journal, issue a correction to whoever received the file, or rebuild the candidate. You are notified, and the record stays in the submission's timeline.

If you have the notice in front of you before any source has indexed it, record it yourself on the Sources tab. Withdrawing it later restores every unit to exactly the status it held.

2.9 Notes

Quick notes with rich text, tags and links to works and projects; personal or shared with a project.

Where notes live

Switch the scope at the top of the Notes page: My notes are private to you; a project scope shows notes every member of that project can read and edit. Folders nest as deep as you like; pin what you use daily.

Templates

New notes start blank or from a template: meeting notes (attendees, agenda, decisions, action items), reading notes (citation, PICO, findings, critique), research idea, or a daily log.

Reading notes pair well with the reader: extract evidence there for anything you will cite, and keep interpretation in the note.

Writing

The note editor shares its engine with the manuscript editor: headings, lists, task lists, tables, quotes, code and links, with Markdown shortcuts (#, -, [ ], >) as you type. Changes save automatically; the header shows when and by whom.

Tags, search and trash

Tag notes freely and search titles, text and tags across your personal notes and every project you belong to. Deleted notes wait 30 days in the trash, where they can be restored or removed for good.

3. Research

Search, evidence, screening, meta-analysis and data.

3.1 The systematic-review workspace

Protocol registration, search strategies with MeSH and PRISMA-S, dual screening with calibration, full text, conflicts, references, risk of bias, GRADE, PRISMA 2020 diagram, grey literature and audit — one sub-app per project.

The stages

Project → Review opens an overview with one card per stage and the live numbers. The navigation follows PRISMA order:

PageWhat you do there
ProtocolFill the PROSPERO / PRISMA-P fields, set eligibility criteria (PICO, inclusion, exclusion, one or two reviewers), record the registration id, lock the protocol; edits after locking are dated amendments; export as Markdown.
SearchBuild one strategy per database from concept groups (title/abstract, controlled vocabulary or all fields), look up MeSH descriptors, keep the generated or hand-edited syntax, record platform, vendor, limits, filters and every run; export the PRISMA-S appendix. Records are imported on the Sources page.
ScreeningTitle/abstract and full-text queues with keyword highlights, notes, keyboard shortcuts (I / E / M, N / P, U undo), blinded co-reviewer votes, AI hints revealed on demand where an agent or the platform has recorded them, and pilot calibration (Cohen's κ between the two most active reviewers).
Full textRecords that passed title/abstract: retrieve open-access full texts, read them, decide with reasons.
ConflictsDisagreements with both votes; an editor records the resolution.
RecordsEvery imported record with its stage, each reviewer's decision, reasons and notes; filter by status, search.
ExtractionEvidence tables with anchored quotations (the Evidence page).
Risk of biasRoB 2 or ROBINS-I per included study; suggestions from an AI agent are labelled and stay suggestions until you confirm them.
GRADEAssemble counts per outcome, run the meta-analysis, grade certainty, insert Summary of Findings and risk-of-bias tables into a manuscript.
PRISMAThe PRISMA 2020 flow diagram from live counts plus the numbers only you know (registers, other sources, reports not retrieved); export SVG or 300-dpi PNG.
Grey literatureChecklist of registers, preprints, theses, conference abstracts, regulatory documents and expert contact with dates and addresses.
AuditEvery recorded action on the project.

With your own AI agent

Connect an agent from the project's Your AI tab and ask it to work the review with you: it can set the protocol, run the search strategy, add and remove records, record screening decisions with reasons, save risk-of-bias and GRADE judgements, assemble and pool the meta-analysis and insert the review tables into a manuscript. Everything it records is labelled and stays advisory: a screening call is a vote until a reviewer confirms it, and a risk-of-bias or GRADE row is a suggestion until someone saves it.

The buttons that ask the platform's own models to do this — “Ask the assistant” on the screening desk, “Propose with the assistant” on risk of bias and GRADE, Pre-fill and Extract on evidence, and the Workflows tab with its systematic-review preset — appear only where an administrator has switched platform AI on. Your connected agent writes the same rows through the same tools either way.

Working without any AI

Every stage runs by hand. Search and import references yourself under Sources and References; screen with the Include / Exclude / Maybe bar (dual screening, conflicts); record risk of bias and GRADE judgements on their pages.

  1. Extraction by hand: Project → Evidence → “Extract by hand”. Pick the study, fill the fields the paper reports (design, population, N, intervention, comparator, primary outcome, key finding, effect measure with estimate and 95% CI, limitations, funding, registration), optionally with page numbers and supporting quotes. The row appears in the evidence table marked “manual” and verified by you.
  2. Pooling: Project → Analysis → Meta-analysis → “Import studies…” pulls effect estimates from the extraction table (generic estimate + CI), or reads one study per row from any dataset with column mapping — event counts, means and SDs, or estimates. Type or paste studies directly too.
  3. Reporting: PRISMA numbers come from your screening decisions; the reporting checklist and the manuscript work the same either way.

AI is optional at every step: it proposes, you decide. Entries made by hand and rows an agent extracted sit in the same table, each marked with its origin, so the audit trail stays complete.

3.3 Meta-analysis

Pool effect sizes from the evidence table or manual entry, inspect heterogeneity and bias, and drop the forest plot into your manuscript.

Starting an analysis

Analysis tab → New meta-analysis. Choose the outcome and the effect measure (odds ratio, risk ratio, mean difference, standardised mean difference, hazard ratio). Studies come from verified evidence units or from a table you type.

Models and heterogeneity

Fixed and random effects (DerSimonian–Laird, REML), with I², τ² and Q reported. Subgroup and sensitivity analyses re-run instantly when you toggle studies.

Plots

Forest and funnel plots render as SVG; insert them into the manuscript as figures whose numbers stay bound to the analysis. Re-running the analysis updates the figure and the bound numbers together.

Reporting

A results paragraph written by an AI agent cites each study and takes the pooled estimate and its interval from the analysis result row as a bound number — never from memory.

The workbench

  • Project → Analysis → Meta-analysis: create an analysis with its effect measure — risk ratio, odds ratio, risk difference, mean difference, standardised mean difference (Hedges' g), hazard ratio or a generic estimate with its CI — then type the studies or paste rows from a spreadsheet (study, year, then the data columns in the table's order).
  • Settings: fixed or random effects (DerSimonian–Laird, Paule–Mandel or REML for τ²), the Hartung–Knapp adjustment, the confidence level, the continuity correction for zero cells, outcome and group labels.
  • Everything recomputes as you type: per-study effects and weights, the pooled estimate with CI, p, I², τ², Q and the prediction interval, a results sentence to copy, and validation warnings (incomplete rows, zero cells, tiny arms, duplicate labels).
  • Views: forest plot (with both fixed and random diamonds), funnel plot with Egger's and Begg's tests and Duval–Tweedie trim-and-fill, leave-one-out, cumulative by year, subgroups with the test for differences, mixed-effects meta-regression with a bubble plot, a metafor R script that reproduces the analysis, and snapshots that freeze a result on record.
  • Export the forest or funnel plot as SVG or 300 DPI PNG. Recorded sandbox runs (code and figures kept as evidence) remain available below the workbench.

3.4 Statistics: tests, sample size and randomisation

Run tests on your datasets, size a study with a power curve, and generate reproducible allocation sequences — each with a methods sentence.

Tests on a dataset

  • Open a dataset and choose Tests: independent t-test (Welch by default, Student's optional), Mann–Whitney U, paired t-test, Wilcoxon signed-rank, one-way ANOVA, Kruskal–Wallis, Pearson and Spearman correlation, chi-squared with Fisher's exact for 2×2 tables, and linear regression with one or more predictors.
  • Assign variables to the slots (scale variables for outcomes, categorical ones for groups; value labels name the groups). Cases with a missing value in any chosen variable are left out, and the active Select Cases filter is respected unless you untick it.
  • Every result gives the statistic, degrees of freedom, an exact p-value, an effect size (Cohen's d, η², r, φ or Cramér's V, rank-biserial r, R²) with confidence intervals where they exist, group summaries or the full table, warnings about small samples or unequal variances, and a sentence ready for the Methods or Results section. Tests are logged in the dataset's history.

Sample size and power

  • Project → Analysis → Sample size & power: two means (Cohen's d or a difference and SD), two proportions, one proportion, correlation, chi-squared and survival (log-rank, Schoenfeld). Set α, power and expected dropout.
  • You get n per group, the total, the achieved power, a methods sentence you can paste, and for two-means designs a power curve showing how power grows with n.

Randomisation

  • Simple, permuted-block or stratified permuted-block allocation for any number of groups; block sizes are multiples of the group count and chosen at random per block.
  • A seed makes the sequence reproducible — keep it in the trial master file. The balance per group (and per stratum) is shown, the sequence can be downloaded as CSV, and a methods sentence is one click away.

3.5 Graphs: publication figures from your data

Bar, line, scatter, box, histogram, Kaplan–Meier, bubble, forest and funnel plots from a dataset or pasted values, styled for journals and exported as SVG or 300 DPI PNG.

Where the data comes from

  • Pick one of the project's datasets and map its variables to the chart (x, y or several y series, group, size, time and event), or paste a small table (CSV or tab-separated, headers in the first row).
  • Grouped summaries (mean, median, count, sum) are computed for you when a categorical x meets a numeric y; the active Select Cases filter of the dataset is respected.

Chart types

  • Bar (grouped or stacked, optional value labels), line (categories or numeric x), scatter (groups and an OLS trend line with R²), box plot (quartiles, whiskers, outliers, mean), histogram (Freedman–Diaconis bins or your own).
  • Kaplan–Meier survival curves per group with Greenwood confidence bands, censor ticks, a numbers-at-risk table and the log-rank p-value.
  • Bubble plot (meta-regression: effect against a covariate, size ∝ weight), forest plot (per-study estimates, weights and pooled fixed/random diamonds computed by inverse variance) and funnel plot with the pseudo-95% funnel.

Style and export

  • Title, subtitle, axis labels, axis ranges, tick rotation, legend position, grid, font size, colour-blind-safe palettes (Okabe–Ito, Viridis, greyscale for print) and journal size presets (single, 1.5 and double column at 300 DPI).
  • Export SVG (vector, best for journals), PNG at 300 DPI, or save both into the project's Files so co-authors can insert the figure into a manuscript. A figure legend is drafted from the mapping for you to edit.

3.6 Data collection forms

Build forms like Google Forms — short answers, choices, checkboxes, dropdowns, linear scales, ratings, yes/no, numbers, dates, times, e-mails — with sections, skip logic and a theme. Every question is a variable of a dataset created for the form (or an existing one), every response a case, and the Responses tab shows live counts, percentages, means and charts.

Create a form

  1. Project → Forms → New form. Name it and pick a template (blank, patient satisfaction, case report form, screening) — a dataset “Form: <name>” is created for you — or choose “Existing dataset” to turn its variables into questions.
  2. Add questions from the panel on the right: each type stores the right kind of variable (a Likert scale becomes an ordinal integer with value labels, checkboxes a text list, yes/no a boolean…).
  3. Edit labels, help text, options and scale anchors on the cards; mark questions required or hidden; reorder with the arrows; add sections to split the form into pages; set “Show only if” for skip logic.
  4. Save changes, then Open form in the header.

Every question is a real variable in Data, so the moment answers arrive you can run frequencies, crosstabs, t-tests or a regression on them — by hand, or by asking your connected agent.

Design and rules

  • Design: theme colour (swatches or any hex), an emoji before the title, tinted or plain background, progress bar, introduction and thank-you message, with a live preview.
  • Rules: anonymous responses, respondent name, optional e-mail, several responses per person or device, shuffled options, close automatically after N responses, show the summary to respondents.

Share links and QR codes

Create as many links as you need — one per clinic, ward or collector — each with its own label, expiry and response cap; revoke any of them at any time. The active link has a QR code to download or print, and an iframe snippet to embed the form in a website. Members can also answer from the Preview tab without a link.

Public submissions are rate-limited per address and per link, validated against the variable types on the server, and never accept passwords or files.

Responses

ViewWhat it shows
SummaryOne card per question: pie or bar chart with counts and percentages for choices, distribution and mean for scales, histogram and statistics for numbers, timeline for dates, latest answers for text; responses per day at the top
IndividualBrowse responses one by one
TableEvery response as a row with who, when and which link

Export CSV or open the dataset in Data for the full statistics engine, graphs and the Python/R runner; a connected AI agent reaches the same dataset over MCP.

3.7 Data: statistics, graphs, Python and R

Every dataset opens in a studio with SPSS-style menus: Data (select, sort, restructure, merge), Transform (recode, standardize, rank, bin, impute…), Analyze (45 analyses from frequencies to Cox regression, factor analysis and clustering), Graphs, a Python/R code runner and an Assistant tab your own AI agent works through. Every result is saved in the Output viewer with the syntax that reproduces it.

The studio at a glance

Project → Data lists the datasets; creating or importing one opens the studio: six tabs (Data · Variables · Output · Graphs · Code · Assistant) under a menu bar (File · Data · Transform · Analyze · Graphs · Code). The Analyze menu opens with a search box and a “Common” shortlist in plain language (compare two groups, predict a yes/no outcome…) above the full catalogue.

TabWhat it holds
DataThe case grid: click a cell and type, Enter moves down, Tab moves right, paste a block from a spreadsheet. Toolbar: undo/redo, add or delete rows, add variable, compute, select cases, pivot, export, history, value labels.
VariablesThe codebook: one row per variable with name, label, type, measure, value labels, missing values and width.
OutputEvery saved result: analyses, code runs and AI answers, pinned ones first. Tables copy as tab-separated text; charts export as SVG or 300-DPI PNG; the syntax that reproduces an analysis is one click away.
GraphsThe chart builder bound to this dataset: bar, line, scatter, box, histogram, Kaplan–Meier, bubble, forest and funnel plots with journal presets.
CodeWrite Python or R against the dataset and run it in the sandbox; the script, printed output, tables and figures are saved together.
AssistantWhere your AI agent works on this dataset: it describes the variables, runs analyses from the catalogue and sandboxed Python or R, and the outputs land in Output. Where an administrator has switched platform AI on, you can also type the question here.

The graphs hub tab moved here: figures belong with the data they draw. The datasets page also has project-wide Outputs and Graphs tabs.

Data files — what comes in and what goes out

Import file (Project → Data) takes CSV, TSV, Excel, SPSS, Stata and R files. All of them are read in your browser: you see the variables, labels and types before anything is stored, and the file itself is never uploaded unless you confirm. SPSS, Stata and R files bring their own dictionaries. File → Export writes the same formats back, instantly, with the dictionary as it stands in the Variables view.

FormatComes in withGoes out with
CSV / TSVColumn headers; types detected from the valuesActive or all cases; value labels in place of codes on request
Excel (.xlsx)The sheet you choose; headers and detected typesData sheet with a frozen header and filters, plus a Variables sheet (the dictionary)
SPSS (.sav, .zsav)Variable labels, value labels, user-missing values (codes and ranges), measurement level, the case weight, datesAll of these, including the weight — the file opens in SPSS with Weight Cases on
Stata (.dta, Stata 13+)Variable labels, value labels, dates and datetimes, strL text; extended missing values (.a–.z) are reported and read as missingFormat 118 (Stata 14+): labels and whole-number value labels; missing codes stay as values (Stata has no missing rules); weighting is a command option in Stata
R (.rds, a data frame)Factor levels as value labels (ordered → ordinal), logicals, Date and POSIXct, haven labels and missing rulesA data frame as saveRDS writes it: factors for coded variables, haven labels where a missing rule exists, Date and POSIXct

Nothing here depends on R or on a server-side converter: the readers and writers are Hikma's own and are checked against SPSS's and Stata's reference reader (ReadStat) and against R itself. Only Stata files older than Stata 13 and R files saved with bzip2 or xz are read on the server, in the isolated sandbox.

Limits: up to 500 variables and the row ceiling shown in the wizard per import; exports write every case up to the same ceiling (use the CSV export for more). The desktop app opens the same files locally and saves them with Save data as….

Data menu — cases

  • Select cases… keeps the cases where a condition is true (age >= 18 AND sex = "F"); the others are greyed out and left out of every analysis unless you untick “Active cases only”.
  • Weight cases… (SPSS Weight Cases) makes a numeric variable a frequency weight: a case with weight 3 counts as three cases in every analysis, table and export, N is the sum of the weights, and cases with a missing, zero or negative weight are left out — the chip in the header says how many. Every graph says whether the weight is applied to it (a scatter-type plot shows each case once, so its points cannot show the weights). Analyses that compute with the weights take any positive weight; those that repeat each case as many times as its weight (Explore, Kaplan–Meier, Cox, ROC, agreement, clustering…) need whole numbers and say so; the runs test refuses. Every weighted result says “Cases were weighted by w (sum of weights N; k cases excluded)” and its syntax starts WEIGHT BY. Scripts — the Code tab and own-data analysis plans — get the weight variable as a column, dictionary.json names it with any user-missing codes, and the run records the weight; the script decides how to apply it. Frequency weights are not survey weights: for a design-based survey analysis use R's survey package in the Code tab.
  • Split file is an option inside every analysis dialog: the analysis is repeated for each level of the split variable, stacked in one output.
  • Sort cases, Remove duplicate cases (by key), Aggregate (one row per group with n, mean, sum, SD…), Restructure wide → long and long → wide, Merge by key and Append cases. Aggregate, restructure, merge and append create a new derived dataset so the original stays intact.

Transform menu — variables

CommandResult
Compute variableA formula over other variables (ROUND, IF, MEAN, DATEDIFF… — see the function list in the dialog), kept up to date when inputs change
Recode into different variableOld value / range / missing → new code, rules top to bottom, unmatched values copied or dropped
Automatic recodeText categories → integer codes with value labels
Standardizez-scores
Rank cases / N-tilesAverage ranks, or quartiles, quintiles… as an ordinal variable
Visual binningEqual width, quantiles, mean ± SD or custom cut points, with generated labels
Count occurrencesHow many of the chosen variables equal a value, fall in a range or are missing
Indicator variablesOne 0/1 dummy per category
Replace missing valuesMean, median, mode, constant, last observation carried forward or linear interpolation
Lag, cumulative sum, differencesIn case order
String function, Date partUPPER/LOWER/TRIM/LENGTH; year, month, day, weekday, quarter

Analyze menu — the statistics engine

Each analysis opens a dialog: pick the variables by role (only suitable ones are offered), set options, watch the live preview, then “Run and save to Output”. Results come as APA-ready tables with statistics, p-values, effect sizes and confidence intervals, a summary sentence you can paste into a manuscript, and charts where they help.

GroupAnalyses
Descriptive statisticsFrequencies, Descriptives, Explore (CIs, percentiles, outliers, normality, histogram, Q–Q), Normality tests, Crosstabs (χ², Fisher, likelihood ratio, residuals, OR/RR/NNT, κ), Missing value analysis
Compare meansMeans by group, one-sample / independent / paired t-tests (Levene, Welch, Cohen's d, Hedges' g), one-way ANOVA with Welch's F and post hoc tests (Tukey, Games–Howell, Bonferroni, Holm, Scheffé), Bootstrap CIs
General linear modelFactorial ANOVA / ANCOVA with interactions and partial η²; repeated-measures ANOVA with Greenhouse–Geisser and Friedman
CorrelatePearson, Spearman, Kendall matrices with heatmap; partial correlations
RegressionLinear (standardized β, VIF, Durbin–Watson, Breusch–Pagan, residual plots), binary logistic (OR, Hosmer–Lemeshow, classification, AUC), Poisson (IRR, offset, overdispersion), multinomial, ordinal (proportional odds, parallel lines), curve estimation (8 models)
NonparametricMann–Whitney, Kruskal–Wallis with pairwise tests, Wilcoxon, sign, Friedman, McNemar, Cochran's Q, chi-square goodness of fit, binomial, runs, one- and two-sample Kolmogorov–Smirnov, median test, Jonckheere–Terpstra
SurvivalKaplan–Meier (numbers at risk, quartiles, survival at times with differences, the log-rank family, restricted means, late entry, landmark), life tables, Cox regression (strata, start–stop rows, shared frailty, changing hazard ratios, adjusted curves, calibration, case checks), parametric models (Weibull, exponential, log-normal, log-logistic) and competing risks (cumulative incidence, cause-specific and Fine–Gray models)
Scale and agreementCronbach's α with item-total statistics, intraclass correlations, Cohen's κ (weighted), Bland–Altman with Lin's CCC
ClassifyDiagnostic accuracy (Se, Sp, PPV, NPV, LRs, DOR with CIs), ROC (DeLong CI, Youden cut-off), k-means, hierarchical clustering with dendrogram, discriminant analysis
Dimension reductionPCA and principal-axis factoring with KMO, Bartlett, scree plot, varimax rotation, communalities

Assumption checks are built in: t-tests and ANOVA report Levene's test and switch the post hoc test when variances differ; Explore and Normality run Shapiro–Wilk, Anderson–Darling and Lilliefors and say which test family to prefer; regressions report VIF, heteroscedasticity and residual normality.

Regression models — the model builder

Linear, logistic, Poisson, negative binomial, gamma and Cox regression share one way of describing a model, and nothing in it is typed as a formula. Choose the outcome and add the predictors; everything else sits in three sections under them, closed until you want them. A section that is in use says so, so a closed section never hides a choice.

WhereWhat you can ask for
PredictorsA categorical predictor takes a reference level. A numeric one takes a shape: a straight line, a curve (a restricted cubic spline with 3, 4 or 5 knots, or a quadratic or cubic polynomial), the log or the square root of the value.
Model: interactionsLet the effect of one predictor depend on another: pick two of the predictors (three for a three-way interaction; the two-way ones it contains are added for you). A spline can interact with a factor.
Standard errors and confidence levelModel-based errors, errors robust to unequal variance (HC3; HC0 to HC2 are there too), or errors clustered on a variable such as patient or site. The same choice carries through to every test, interval and estimate after the model. The confidence level is 95% unless you choose 90% or 99%.
After the modelFor a factor: its adjusted means (or probabilities, rates) with comparisons between its levels: every pair, each against the reference, a trend, or a contrast of your own, adjusted by Tukey, Šidák, Bonferroni, Holm or the false discovery rate. For a number: its slope, or predictions at values you give, with a plot of its effect. “Within each level of” repeats this inside the levels of a second predictor, which is how an interaction is read. Average effects puts every effect on the outcome's own scale: a difference in mean, in risk or in rate.

The output adds what the model calls for. With an interaction or a curve, a table tests each term as a whole (Type II tests, so a main effect is not judged at the reference level of something it interacts with), a second table asks whether each curve is needed, and the interaction is opened up for you: the means of one factor within the other with their comparisons, or the slope of a number in each group, and a plot with confidence bands. Robust or clustered errors are named in the tables and in the summary sentence.

For this questionChoose
Risk ratios or risk differences instead of odds ratiosBinary logistic regression → Report: risk ratios (modified Poisson with robust errors always fits; log-binomial is the classical model) or risk differences (identity link)
A skewed positive outcome such as cost or length of stayGamma regression: ratios of means on the outcome's own scale, with nothing to back-transform
Counts more spread than Poisson allowsNegative binomial regression, or Poisson with “Quasi-Poisson” ticked, or Poisson with robust errors
Patients measured in several sites, or several rows per patientStandard errors: clustered, with the site or the patient as the cluster variable. With fewer than about 30 clusters the output warns that such errors run small.
Does the treatment work differently in men and women?Add both as predictors, add their interaction, and read “Comparisons of treatment within sex” and the plot
Is the effect of age a straight line?Enter age as a spline and read “Is the curve needed?” and the effect plot

Every number of this part of the engine is checked against R (lm, glm, coxph with the sandwich, car and emmeans packages) and against Python's statsmodels on five datasets that ship with the tests. Robust errors follow sandwich; clustered errors use the same small-sample factors as R and Stata; comparisons follow emmeans.

A saved run writes its whole specification as one syntax line (ANALYZE … interactions=[[arm, sex]] vcov=robust after={…}), and the session's R script writes the same model with lm or glm, sandwich, car::Anova and emmeans, so a colleague can repeat it either way.

Comparing groups and repeated measurements

The General linear model group of the Analyze menu holds the analyses of variance. Each one checks its own assumptions, reports the size of every effect, compares the groups for you, and opens with a reading in plain words.

AnalysisUse it whenWhat you get
One-way ANOVAOne grouping factor, a numeric outcomeF and Welch's F, Levene's test, η² with its interval and ω², post hoc comparisons (Tukey, Games–Howell, Bonferroni, Holm, Scheffé, LSD). Planned contrasts: a trend across the groups, each group against a control (Dunnett), or weights of your own.
Factorial ANOVA / ANCOVASeveral grouping factors, with or without covariates to adjust forUp to five factors with their interactions; Type III tests (as SPSS and SAS) or Type II; partial η² with its interval and ω² for every effect; Levene's test; for an ANCOVA, whether the covariate's slope is the same in every group; adjusted means of every factor with all pairs compared; each interaction opened up; profile plots.
Repeated-measures ANOVAThe same cases measured several times, with or without groupsOne or two within-subject factors laid out over the columns, any grouping factors; Mauchly's test with the Greenhouse–Geisser and Huynh–Feldt corrections; the multivariate test; partial and generalized η²; the means with every pair compared; the change inside each group when the groups change differently; a profile plot.
Mann–Whitney, WilcoxonTwo groups or two measurements, a skewed or ordinal outcomeThe test, the exact p-value for small samples without ties, rank-biserial r, and the Hodges–Lehmann difference with its 95% interval.
Kruskal–Wallis, FriedmanThree or more groups or measurements, a skewed or ordinal outcomeThe overall test with its effect size and the pairs that differ (Dunn's test after Kruskal–Wallis), Holm-adjusted.
  1. Two within-subject factors: choose the measurements in the order where the second factor changes fastest (visit 1 condition 1, visit 1 condition 2, visit 2 condition 1…).
  2. Open “Two within-subject factors”, name the first factor and give the number of levels of the second.
  3. The output starts with a table of how the columns were read. If it does not match your data, reorder the measurements.

When a dialog cannot run yet it says what it still needs (“Still needed: Dependent variable, Fixed factors”). Where a level is typed, such as a reference level or a control group, the dialog suggests the variable's own values.

Factorial ANOVA results saved before version 0.7.1 of the engine tested the main effects of a model with an interaction at a reference level, under the heading Type III. Run them again to get the Type III tests that SPSS, SAS and R report.

Tables of counts, risks and rates

These analyses answer the questions of a clinical or epidemiological study from counts: how much more often an outcome occurs in one group than another, whether that holds across strata, how fast events occur, and whether two treatments are close enough to be called the same.

AnalysisUse it whenWhat you get
CrosstabsTwo categorical variablesThe table with χ², the likelihood-ratio test and Fisher's exact test (for 2 × 2 tables, and for larger tables whenever counts are small). For a 2 × 2 table: the risk in each group, risk ratio, risk difference, number needed to treat, odds ratio (with the exact one), and attributable fractions, read as exposed against reference for the event. With a layer: the Mantel–Haenszel common odds ratio and risk ratio, the Cochran–Mantel–Haenszel test, and the Breslow–Day test that the odds ratio is the same in every layer. With three or more ordered groups: the Cochran–Armitage test for trend.
Binomial testOne proportion against a valueThe exact test and the Wilson, exact (Clopper–Pearson), Agresti–Coull and Wald intervals, with which one to report.
Incidence rates and rate ratiosEvents counted over follow-up timeThe rate per 1,000 (or any number of) person-years in each group with its exact interval, and each group's rate ratio and rate difference against the reference group. Standardise for a second variable (age band, sex) to compare groups of different make-up: directly standardised rates, the Mantel–Haenszel rate ratio, and the SMR (observed events against those expected at the reference rates).
Ordinal regressionAn ordered outcome (mild, moderate, severe)One odds ratio per predictor under proportional odds, with a check of that assumption for each predictor (Brant's test) and the odds ratio at every cut-point of the outcome, so you can see where an effect changes along the scale.
Conditional logistic regressionA matched case-control study, 1:1 or 1:manyOdds ratios that compare each case with the controls of its own set, so the matching factors drop out. Interactions, curved effects, a test of every term and comparisons between levels, as in the other regressions. The output shows how the sets are made up and which carry no comparison.
Equivalence and non-inferiorityThe aim is to show that two groups do not differ by more than a margin, or that one is not worse than the otherTwo one-sided tests against the margin, the interval that answers the question (90% at α = 0.05) beside the usual 95% one, a plot of the interval against the margin, and a verdict in words: equivalent, not equivalent, or inconclusive. For two groups, paired measurements, one group against a value, or two proportions; the margin in the outcome's units, in standard deviations, or as a ratio such as 0.80 to 1.25.
  1. For a 2 × 2 table, check the title of the risk table: it says which group is compared with which, and for which outcome (“yes against no, for event = 1”).
  2. If that is not the direction you want, open “2 × 2 tables: how to read the risks” in the dialog and give the event value or the reference group. The dialog suggests the variable's own values.
  3. Risks, risk ratios and risk differences suit a cohort, a trial or a survey. In a case-control study only the odds ratio has a meaning.

A test that finds no significant difference has not shown that two groups are the same: the study may simply be too small. To claim “no meaningful difference”, decide beforehand how large a difference would matter and run Equivalence and non-inferiority with that margin.

In a matched study do not add the matching variables (the age band, the sex, the clinic) as predictors of the conditional logistic regression. They are identical within every set, so they cannot be estimated; the matching has already held them constant.

Measurements that belong together

When the same people are measured several times, or people sit inside clinics, schools or families, the rows are alike within a person or a site. Treating them as independent makes every test too sure of itself. These three analyses allow for it. They need the data in long format, one row per measurement; Data → Restructure turns a wide file (one column per visit) into a long one.

AnalysisUse it whenWhat you get
Linear mixed modelA numeric outcomeEach person (or site) has its own level and, if you ask, its own rate of change; a second grouping such as the clinic above the patient. Tests of the predictors with Satterthwaite degrees of freedom, how much people differ, whether they do, each person's predicted level, a plot of everyone's measurements under the model's average line, and adjusted means with comparisons.
Mixed model for a yes/no outcome or a countA yes/no outcome or a count, and the question is about one personOdds or rate ratios for a given person, how much people differ (the median odds ratio), a test of whether they do, each person's predicted level.
GEEAny of the three outcome types, and the question is about the average of the whole groupPopulation-averaged effects with robust standard errors, corrected when the groups are few; the working correlation chosen with a table that compares all of them.
  1. Choose the outcome and the predictors as in any regression.
  2. Under “Rows that belong together” choose the variable that marks the person or the site.
  3. If people may change at different rates, open “What differs between the groups” and tick the time variable.
  4. Read “What this means” first: it says how much of the variation lies between people and what each predictor does once that is allowed for.

A mixed logistic model and a GEE answer different questions and give different odds ratios from the same data. The mixed model's are for one person and lie further from 1; the GEE's are for the group as a whole. Report the one that matches your question and say which it is.

Mixed-model results saved before 10 November 2026 used a faulty REML criterion (variances about 1% off), and a GEE's QIC had the wrong penalty. Run them again for the corrected values.

Time until an event

Use these when the outcome is how long something takes and some people have not had it by the end of their follow-up: death, relapse, discharge, a device failing. Each person needs a follow-up time and a variable that says whether the event happened. People who had not had it are censored: they count for as long as they were followed, and no longer.

AnalysisUse it whenWhat you get
Kaplan–MeierYou want the curve, or to compare groupsThe curve with the numbers at risk under it; the time by which a quarter, half and three quarters have had the event; survival at the times you choose with the difference between groups; a test of the curves (log-rank, or one that weighs early or late differences more); the average event-free time; a hazard ratio per group.
Life tableFollow-up is reported in intervals, or all you have is counts per intervalFor each year or month: how many entered, were withdrawn and had the event, the share still event-free at its end, and the hazard.
Cox regressionSeveral predictors at onceA hazard ratio per predictor, each allowed for the others; a check that each ratio holds over the whole of follow-up; survival curves adjusted for the other predictors; how well the model predicts by a chosen time; the cases that pull hardest on a result.
Parametric survival modelYou need a predicted median, or to look past the end of follow-upFour shapes of hazard fitted and compared (Weibull, exponential, log-normal, log-logistic); effects as time ratios; observed against fitted curves.
Competing risksSomething else can happen first and rule the event outThe share who have had the event by each time, the share who had the other outcome first, a test between groups, and each predictor's effect on the rate of the event and on the share who end up with it.
  1. Choose the follow-up time and the variable that says what happened; say which value is the event if it is not 1 or Yes.
  2. Add a grouping variable (Kaplan–Meier) or predictors (Cox, parametric).
  3. Read “What this means” first. Then look at the numbers at risk: a curve is only as firm as the number of people still under it.
  4. Open the optional sections only when your data need them: people who join late, a person with several rows, groups formed during follow-up.
Your dataWhat to do
People join the study at different points of the time scale (age as the time scale; enrolled some time after diagnosis)Give the time each person joined under “When people join late”. They are counted as at risk only from then.
A value changes during follow-up (a complication sets in, a treatment starts)One row per period for each person, each with its start and end, the event on the last row; give the start under “Row starts at” and the person under “Person”. Do not carry the final value back to time 0: someone must survive long enough to get it, and that waiting time would be credited to it.
Groups are formed during follow-up (responded by three months)Set a landmark time. Only people still event-free then are analysed, grouped by what had happened by then.
People come from centres, clinics or familiesFor a few large centres: stratify by centre. For many small groups: “Groups that share an unmeasured factor” (a shared frailty), which also tells you how much the groups differ.
A hazard ratio does not hold over follow-up (the check says so, or the curves cross)Let it change: smoothly with time, or with a value of its own before and after times you choose. Or report the average event-free time from Kaplan–Meier, which needs no such assumption.

A hazard ratio above 1 means the event comes sooner. A time ratio (parametric models) above 1 means it comes later. They point in opposite directions for the same effect; the output says which one it is showing.

Survival results saved before 10 November 2026 should be run again if you report any of these: the interval of a median (its limits were printed in reverse order), the interval of a restricted mean (about 6 to 9% too narrow), numbers at risk at a time point (people whose time was exactly that point were left out), the test of proportional hazards (now the one R prints since 2019), or a hazard ratio allowed to change with time (its test against a constant ratio was not valid when times were tied). Hazard ratios, survival percentages and log-rank tests were right.

Gaps in the data, and treatments that were not randomised

Two problems that a plain comparison hides. Cases with a missing value drop out of a model without a word, and the cases that are left may not resemble the ones that went. And when people were not assigned to a treatment at random, the treated differ from the others before treatment starts, so a difference in outcome is not yet an effect.

AnalysisUse it whenWhat you get
Missing value analysisBefore any model, if some values are missingHow much is missing and in which combinations; a test of whether the gaps are completely at random; for each variable, what goes with its gaps; a plain statement of whether leaving cases out is safe.
A model with multiple imputationLeaving cases out would bias the result or cost too many casesEach gap is filled several times from the other variables, the model is fitted to every filled-in copy and the results are combined. Pooled coefficients and tests, a comparison with the complete cases, a look at what was imputed, and a what-if for gaps that are not at random.
Propensity scoreA treatment or exposure that people were not assigned to at randomA model of who got the treatment; the groups made alike by weighting or matching; balance before and after; the effect with its interval; a second, doubly robust estimate; for a yes/no outcome a risk ratio; for a time to an event weighted curves and a hazard ratio.
Instrumental variablesSomething you did not measure drives both the exposure and the outcome, and you have a variable that moves the exposure and nothing elseThe effect estimated through the instrument, beside the ordinary regression; how strongly the instrument moves the exposure; whether the exposure is confounded at all; whether several instruments agree; an interval that stays valid when the instrument is weak.
Difference in differencesA policy, programme or system reached one group at a known time and not another, and you have the outcome before and afterEach group's change and the difference between them, with errors clustered on the units followed; with several periods the gap between the groups period by period and a test that they moved in parallel before the change.
Regression discontinuityWho gets something is decided by a score and a threshold: a programme from a risk score of 50, a drug below a blood valueThe jump of the outcome at the threshold, corrected for curvature and with a robust interval; the same with a narrower and a wider window; a picture; a test that cases do not pile up on one side; other variables that should show no jump.
MediationYou know that an exposure affects an outcome and want to know how: through an intermediate variable, or by other routesThe effect through the mediator, the effect by other routes, the total and the share mediated, with bootstrap intervals; the two models behind them. Outcome and mediator can each be a number or yes/no.
  1. Run Missing value analysis on the variables of your model. If it says the gaps do not matter, go on as usual.
  2. If they do matter, open “Missing values” in the model's dialog and choose to keep the cases. Add helper variables that go with the gaps.
  3. For a treatment that was not randomised, open Propensity score. Choose the treatment, every covariate measured before treatment that could drive both treatment and outcome, and the outcome.
  4. Choose whom the answer is about, then read the balance table before the effect: an effect from groups that are still unbalanced means little.
The effect in whomThe question it answersChoose it when
EveryoneWhat if all were treated against noneThe treatment could sensibly be given to anyone in the data.
The treatedWhat the treatment did for those who got itYou are judging a treatment as it was actually used.
Those who could have gone either wayThe effect where there was a real choiceThe groups differ a lot. These weights balance the covariates exactly and never become extreme.
The treated, by matchingThe same as the treated, in pairs or sets you can listYou want a matched sample to show. Treated cases with no counterpart are left out, and the output says how many.

A propensity score makes the groups alike only on the covariates you gave it. It does nothing about what was not measured. Say so when you report it, and give the largest standardised difference after adjustment.

A propensity score deals with confounders you measured. An instrument deals with those you did not, but only if it really touches the outcome through the exposure alone. No test can show that for a single instrument: you have to argue it from how the data arose. A random encouragement or a lottery is a strong argument; a convenient variable that happens to correlate with the exposure is not.

Look at the first-stage F before the estimate. Under about 10 the instrument is weak: the estimate drifts towards the ordinary regression and its usual interval is too narrow. Report the Anderson–Rubin interval then, even if it is very wide.

A difference in differences is the effect of the change only if the two groups would otherwise have moved in parallel. Collect several periods before the change if you can: the period-by-period table then shows whether they did. Give the unit that is followed over time (the clinic, the region), or the interval comes out too narrow.

A regression discontinuity is only as good as the rule behind it. If people or their doctors could nudge the score over the line, cases pile up on one side and the two sides are no longer alike: the density check looks for exactly that. The result holds for cases near the threshold and says nothing about those far from it.

Mediation asks more of the data than any other analysis here. Even when the exposure was randomised, the mediator was not: anything that drives both the mediator and the outcome is taken for an effect through the mediator. Add such variables as other predictors, and say in your report that the result rests on having measured them.

If the weighted estimate and the doubly robust estimate disagree, one of the models is wrong. Enter the strongest covariates as curves, add an interaction, and look at the balance again.

Imputed models run before 10 November 2026 should be run again if a variable with two categories had gaps: those gaps were filled at random, whatever the other variables said. The column headed FMI also held a smaller quantity than its name says.

Code — Python and R

The Code tab runs your own script in an isolated sandbox (no network, three-minute limit) against the active cases. Python has pandas, numpy, scipy, statsmodels, lifelines, scikit-learn, seaborn/matplotlib and pingouin; R has survival, ggplot2, lme4, metafor, meta, pROC, car, emmeans, psych, irr, nnet and ordinal.

Python: numbers with hikma.result, tables with hikma.table, figures with hikma.figure(name, fig, caption)

import pandas as pd
d = pd.read_csv(DATA_PATH + "/data.csv")
hikma.result("n", len(d), label="Cases")
hikma.table("means", d.groupby("arm")["sbp"].mean().reset_index(), "Mean SBP by arm")

R: hikma$data() reads the cases; hikma$result / hikma$table / hikma$figure report

d <- hikma$data()
m <- coxph(survival::Surv(time, status) ~ group, data = d)
hikma$table("hr", as.data.frame(summary(m)$conf.int), "Hazard ratios")
hikma$figure("km", function() plot(survival::survfit(survival::Surv(time, status) ~ group, data = d)), "Kaplan–Meier")

Templates in the tab give you a starting script for descriptives, tests, mixed models, Cox models and plots. Every run is saved to Output with its script, so a reviewer can rerun it.

Analysing with AI

Ask your connected agent (project → Your AI) about the dataset in plain language — “does systolic blood pressure differ between arms after adjusting for age?”. Over MCP it describes the dataset and its variable dictionary, runs analyses from the catalogue, writes and runs sandboxed Python or R for what the catalogue cannot express, and reads the results back; every output lands in the Output viewer with the syntax that reproduces it, so its interpretation can be checked number by number.

The Assistant tab's own question box — type a question, the platform plans 1–6 steps, runs them and writes the interpretation from the returned results only — needs platform AI switched on by an administrator. In the default mode the tab explains where to ask instead.

  1. Describe what you want to know, naming the project and the dataset.
  2. The steps are planned against the variable dictionary, preferring analyses from the catalogue over hand-written code.
  3. Each step runs on your real data in the sandbox; nothing is estimated.
  4. The interpretation must come from the returned tables — every number copied, with caveats — because that is all the results contain.
  5. The answer, its steps and their tables are saved to Output and listed under Earlier questions.

From an MCP client (Claude, Codex, Cursor…), the same power is available through describe_dataset, list_analysis_catalog, run_dataset_analysis, transform_dataset, dataset_cases, run_dataset_code and list_dataset_outputs.

From Output into the manuscript

Every saved output has “Insert into manuscript”: choose the manuscript, what to include (the summary sentence, the tables, the charts) and where; the tables arrive as real tables, the charts as figures with captions, all as tracked changes attributed to whoever inserted them — “Claude Code” or whatever your connected agent is called — so co-authors can accept or reject them. Meta-analysis results, risk-of-bias and Summary-of-Findings tables insert the same way from their own pages.

Identifier scan and de-identification

Every dataset is scanned when it is created or imported: variable names, labels and a sample of values are checked for direct identifiers (names, e-mail addresses, telephone numbers, postal addresses, national or hospital numbers, dates of birth, device identifiers). A red or amber banner in the studio lists the flagged columns with the reason. Remove or recode them (Variables tab, Transform › Recode), re-scan, or record an attestation that the data are de-identified — the attestation is kept with the dataset's history with your name and the time. Study codes issued by the project are flagged as “possible” only so you can confirm them.

Graphs — from the dataset, saved as outputs

The Graphs tab (and Graphs › New graph) builds a figure from the dataset's own variables: bar (counts or a summary by category, stacked), line, pie, mean ± SD/SE/95% CI, scatter with trend line, box plot, histogram, Kaplan–Meier with log-rank, bubble, forest and funnel plots. Map the variables and the graph draws at once; style it (title, axes, palette, journal size) and press “Save graph” — it is stored in Output as a chart result with its settings, so you can reopen it with “Edit graph”, export it as SVG or 300-DPI PNG, and copy it to project files. The project's Data › Graphs gallery lists every saved graph across datasets; “New graph” always starts from a dataset. Numbers that are not in any dataset can still be graphed from the gallery's “typed values” panel.

Analyses save their own charts too (histograms with normal curve, Q–Q, error bars, ROC, survival curves, forest plots), and the Analysis hub's “Run an analysis” opens the chosen procedure on the chosen dataset in one step.

Output viewer

  • Pin what matters; rename outputs; delete drafts.
  • Copy any table as tab-separated text for Word or Excel; export any chart as SVG or 300-DPI PNG.
  • Analyses carry a Hikma syntax line (ANALYZE …) — the exact specification that produced them.
  • The datasets page → Outputs lists results across all datasets of the project.

3.8 Slides: presentations in the project

Presentations with Hikma's own slide editor: layouts, text with bullets, pictures, tables, live charts, shapes, themes, speaker notes, a slide sorter and a presenter view — saved as ordinary .pptx files in the project that PowerPoint, Keynote and LibreOffice open. A manuscript becomes an editable deck in one step. The same Slides runs in the desktop app on files on your computer.

The editor at a glance

Project → Slides lists the project's presentations; New presentation creates a .pptx in the project's files, and a .pptx uploaded under Files appears here too. A presentation opens with the slide list on the left, the slide in the middle and its properties on the right: the selected element's position, size, turn, font and colours; the slide's layout, background, hidden flag and speaker notes; the presentation's theme. Save writes the file back to the project; an edit is also saved twenty seconds after the last change.

Do thisTo
Click · Shift+click · drag on the empty slideSelect one, add to the selection, select everything the drag touches
Drag · the eight handles · the round handle aboveMove (pink guides show what it lines up with; Alt turns them off), resize (Shift keeps the proportions), rotate (Shift: 15° steps)
Double-click or Enter · EscEdit a text box, a shape's label or a table cell in place · leave it
Tab · Shift+Tab in a listIndent a bullet · bring it back
Delete · arrows · Shift+arrowsRemove · nudge by 1 · nudge by 10
Ctrl+C / X / V · Ctrl+DCopy, cut, paste (a table copied in Excel, Sheets or Word pastes as a table, a picture as a picture) · duplicate
Ctrl+Z / Y · Ctrl+MUndo and redo · a new slide
F5 · Shift+F5Present from the first slide · from this one

Slides, text, tables, charts and pictures

New slide offers seven layouts: title, title and content, two content, section header, figure with caption, table, blank. Changing a slide's layout keeps what was typed in its title and body. The toolbar inserts a text box or a bulleted list, a picture, a table (rows and columns are added and removed in the right pane), a chart and shapes (rectangle, rounded rectangle, ellipse, triangle, diamond, line, arrow).

  • Chart opens a small grid: type the numbers or paste them from Excel or Sheets, choose bars, stacked bars, horizontal bars, lines, a pie or a scatter, and the chart is drawn as you type. Double-click a chart to change it. It stays a live chart in Slides and is stored in the .pptx as a clean picture, so PowerPoint shows it as you made it.
  • The text bar sets bold, italic, underline, size, colour, bullets, numbering and alignment: on the selected words while a box is being edited, on the whole box when it is only selected.
  • Arrange aligns and distributes the selection, changes what is in front, and rotates by quarter turns.
  • Theme changes the colours and fonts of the whole presentation; Edit theme… makes your own. Everything that still uses the old theme follows; colours you chose by hand stay.
  • The grid button shows every slide at once for reordering by drag; the eye hides a slide from the show without deleting it.

Presenting and exporting

Present fills the screen with the slides. Arrows, Space or a click move on; right-click goes back; a number then Enter jumps to that slide; B blanks the screen; L turns the pointer into a laser dot; Esc ends the show. Move the mouse for the controls: Presenter view opens a second window with the current slide, the next one, your notes at a size you choose, a clock and a timer, and its keys and buttons drive the show.

  • Export → PDF writes one page per slide at the slide's own size.
  • Export → This slide as a picture saves a 1920 × 1080 PNG.
  • Export → Print… prints on paper sizes with real text; choose landscape in the dialog.

The exported PDF's pages are pictures of the slides: sharp on a projector and on paper, but the text in them cannot be selected. Use Print… when you need selectable text.

A presentation from a manuscript

In a manuscript, choose Export, pick Slides as the format and then Edit the slides…: the slide builder shows one section per heading, its sentences as points, its tables and figures on slides of their own, references at the end. Edit the outline there, then choose Edit in Slides: the deck is created in the project and opens in the editor, where every slide can be changed like any other. The first slide says that the deck is an unchecked working copy of the live manuscript, as every working export does; remove it when the content has been through its checks.

Citation numbers in the points are the manuscript's own, so the reference slides match the text.

Files, other programs and the desktop app

A presentation is an ordinary .pptx. Files made in Slides come back exactly as they were. Files from PowerPoint, Keynote, Google Slides or LibreOffice open as those programs draw them: placeholders take their position, sizes and bullets from the layout and master, colours come from the theme, the master's artwork is copied onto each slide and locked, SmartArt opens as the shapes it is drawn with, and drawn shapes Slides has no kind for are kept as pictures of themselves. What cannot be kept — native PowerPoint charts, video and sound, animations and transitions — is listed under the editor when the file opens.

The desktop app has the same Slides for files on your computer, without an account: Slides in its sidebar. There, an analysis output in Data and a range or chart in Sheets go to a presentation with Send to Slides and Insert in Slides.

3.9 Sheets: workbooks in the project

Spreadsheets with Hikma's own engine: 230 functions that recalculate as you type, number and date formats, conditional formatting, data validation, charts, pivot tables and named ranges — saved as ordinary .xlsx files in the project that Excel and LibreOffice open unchanged. The same Sheets runs in the desktop app on files on your computer.

The workbook at a glance

Project → Sheets lists the project's workbooks; New workbook creates an .xlsx in the project's files, and an .xlsx uploaded under Files appears here too. A workbook opens with a toolbar (undo, font, bold/italic/underline, colours, alignment, wrap, merge, number format, decimals, sort, filter, freeze, find, Rows, Columns, Insert, Format, Data), a name box and formula bar, the grid, and sheet tabs with Add sheet and a Sheet menu (rename, move, colour, delete). Save writes the file back to the project; 20 seconds after your last change it saves on its own.

KeysWhat they do
Enter · Tab · Shift+Enter · Shift+TabCommit and move down, right, up, left
F2 · Esc · DeleteEdit in place · cancel · clear
Ctrl+arrows · Ctrl+Home/End · Shift+clickJump across data · corners · extend the selection
Ctrl+C / X / VCopy, cut, paste — as cells inside Sheets, as tab-separated text with Excel, Word and LibreOffice; pasted text that starts with = becomes a formula
Ctrl+Z / Y · Ctrl+B / I / UUndo, redo · bold, italic, underline
Fill handleDrag the small square at the selection's corner to continue a series or copy formulas with relative references

Formulas and functions

A cell that starts with = is a formula: cell and range references (A1, B2:D9, A:A, 3:3), other sheets (Sheet2!A1, 'Site data'!B2:B9), absolute references ($A$1), names (=AVERAGE(Ages)) and 230 functions — maths, statistics (STDEV.S, T.TEST, NORM.DIST, CORREL, SLOPE…), text, dates (DATEDIF, NETWORKDAYS, EOMONTH…), lookups (VLOOKUP, XLOOKUP, INDEX/MATCH, FILTER, UNIQUE, SORT), logic (IF, IFS, SWITCH, IFERROR) and information. The formula bar completes function names (Tab accepts) and shows the signature while you type. Recalculation is incremental and immediate; a circular reference shows #CYCLE! in the cells involved.

Results agree with Excel: 195 functions were written by Sheets and recalculated by LibreOffice in the compatibility audit with the same answers; the statistical functions are computed to machine precision.

Formats, conditional formatting and data validation

The number format list has General, numbers, thousands, percent, scientific, currency, dates, times, text and Custom (any Excel format code). Typed input implies a format once: 12.5% becomes 0.125 shown as 12.5%, 2026-03-15 a date.

  • Format → Conditional formatting… adds rules to the selection: cell value is (eight operators), text contains, top/bottom N, duplicate values, a formula that is TRUE, two- or three-colour scales; fill, text colour and bold. Rules are listed per selection and removable; Excel keeps them.
  • Data → Data validation… restricts a range: a list of options (typed, or a range) with a dropdown arrow on the cell, whole or decimal numbers, dates, text length with operators, a custom formula, blank allowed or not, and your own message. A refused value stays in the editor with the message.
  • Data → Sort… (by a column, header row kept, formulas follow their row) and Filter… (tick the values to show) work on the current block; Freeze panes… keeps rows and columns in view.

Charts, pivot tables and named ranges

  • Insert → Chart…: bar, line, pie, scatter, histogram, box plot from a range (first row as series names, series in columns or rows). The chart sits on the sheet, follows the cells, moves by its title bar, resizes at its corner, saves as a PNG, and goes into the .xlsx as a picture at its place — the definition travels in the file so it stays live here.
  • Insert → Pivot table…: row field, optional column field, value field and count/sum/mean/median/min/max/SD with totals; the summary becomes a new sheet (re-run for a fresh one).
  • Insert → Named range…: names for ranges, usable in formulas and saved as Excel names.

Files: .xlsx, CSV, print

Workbooks are ordinary .xlsx files: values, formulas with their results, formats, fonts, fills, alignment, borders, widths, merges, frozen panes, validation and conditional formats as Excel stores them; Sheets' charts also as pictures. CSV exports the sheet or the selection with your separator and decimal mark (semicolon and comma for European Excel); Print puts the used range or the selection on paper or in a PDF through the browser's print dialog.

The desktop app runs the same Sheets on files on your computer (Sheets in the sidebar, or double-click an .xlsx) and adds two hand-offs: Analyse in Data opens the block as a dataset in Data, Insert in Writer places it as a table in the open document. Signed in, the desktop app shows the project's workbooks under Projects → Sheets as well.

4. Writing and submission

The manuscript editor, checks, journals and reviewer responses.

4.1 The manuscript editor

A collaborative editor built for papers: templates, tracked changes, object citations, figures, tables, equations, comments, versions and durable authorship — with your own AI agent writing into it over MCP.

Creating a manuscript

  1. Manuscripts tab → New manuscript.
  2. Pick a template: original research (IMRaD), systematic review, meta-analysis, case report, protocol, letter, review, thesis chapter, grant section and more. Templates seed the sections and the reporting checklist.
  3. Or import a Word document: headings, lists, tables, figures and numbered citations are recovered and re-linked to works in the project where possible.

Front matter

Title, running title, authors with CRediT roles and affiliations, structured or free abstract, keywords, disclosures (funding, conflicts, ethics, data availability), target journal with word and figure limits, page setup and citation style live in the Front matter panel and are rendered by exports and checked by the reporting checks.

Writing together

  • Real-time collaboration: everyone sees everyone's cursor; nothing is lost on conflicting edits.
  • Tracked changes with author and time for every insertion and deletion; accept or reject one by one or per author. Turn tracking off for your own drafting.
  • Comments anchored to text with replies, resolution and @mentions; suggested edits from reviewers.
  • Slash menu (/) for headings, lists, tables, figures, equations, page breaks and captions (AI actions join it where an administrator has switched platform AI on).
  • Find and replace with case, whole-word and regular expressions; deleted text never matches.

Citations, figures, tables and equations

Citations are objects pointing to works in the project; the reference list and inline labels are rendered from those records in the chosen style. Figures carry captions, widths, alignment and panels (add a second image from the figure's toolbar: it becomes panel B, lettered and laid out in columns, with one caption); tables have borders, shading, alignment, sorting and text-to-table. Equations have their own editor (Insert → Equation, or Alt+=): a palette of structures, Greek, operators and statistics templates, LaTeX underneath and a live preview; click an equation to edit it. Display equations can be numbered and cited as “Equation 2”, and they export as native Word equations, typeset PDF and TeX.

Side panels

PanelUse
OutlineHeadings; drag to reorder sections
Front matterMetadata described above
StatisticsWords, characters, readability, estimated pages against the journal limit
AssistantWhat your connected AI agent is doing in this manuscript, and how to connect one; its text arrives as tracked changes
ChangesEvery pending change with accept and reject
CommentsThreads with resolution
AuthorsWho wrote what: colour the manuscript by author and see every contributor's share
ChecksReporting and house-style checks
VersionsNamed snapshots; compare two as a report or as a redline in the editor (insertions underlined, deletions struck); restore, or bring a version back as suggestions that the Changes panel reviews
NotesScratch notes for this manuscript

Who wrote what

Every run of text remembers who wrote it. Type a sentence and it carries your name; accept a colleague's suggestion and the text keeps *their* name, not yours; accept a paragraph your connected agent drafted over MCP and it stays attributed to that agent by name — “Claude Code”, “Cursor”, whatever the key is called. Attribution is part of the manuscript, so it survives the change being accepted, the page being reloaded and the work being picked up months later — which is what a tracked change alone cannot tell you, because it disappears the moment someone accepts it.

  1. Open the Authors panel on the right.
  2. Colour by author paints the manuscript, one stable colour per person; hover any sentence to see who wrote it and when.
  3. The contributors list gives each person's share in words and paragraphs, and how much of theirs is still waiting as a proposed change. Click a contributor to keep only their text coloured and dim the rest; click again for everyone.
  4. At the cursor names the author of the text you are standing on as you move through the document.
  • Not recorded covers text written before this manuscript kept attribution, or brought in by an import — it is counted separately rather than credited to whoever touched it last.
  • A sentence still shown as a proposed change is marked *still a proposal*: it names who proposed it, and becomes permanent when the change is accepted.
  • Attribution never leaves the editor. Exports, journal packages and submissions carry the text alone — no names, no colours.

Authorship answers who typed a sentence. For where its facts come from — the cited work, the page, the quotation, the analysis behind a number — use the Sources panel.

Versions

Save a named version before big edits or a submission. Compare any two versions as a diff, or restore one (the current state is saved first, so nothing is lost).

Export

Word (with revision marks and comments), PDF, HTML, Markdown, LaTeX (with .bib and figures), JATS XML (see below) and PowerPoint for talks. Exports are blocked while a citation points to an unknown or retracted work or a number points to deleted evidence, unless you override with a reason. An AI-use statement is appended: composed from the record where the platform's own AI was used, and otherwise a line for you to complete with the agent and model you used.

JATS XML

Export → JATS XML writes the manuscript as a NISO JATS 1.3 Journal Publishing article — the XML journal production systems, PubMed Central and repositories read. It downloads as a zip: manuscript.xml, the figure files it references under figures/, and manifest.json, which says what the file is (an UNCHECKED working copy, or the release candidate it was rendered from), lists every file in the zip with its sha256, records the DTD validation and names anything the format did not carry. Your connected agent gets the same file and the same record with render_working_export or render_release (format jats).

  • Front matter: the title and running title; authors with ORCID iDs, degrees, CRediT roles and affiliations; the corresponding author; the abstract — a structured abstract becomes one titled section per heading — keywords and funding. Conflicts of interest and author contributions are author notes. Acknowledgements, data availability, ethics, consent, registration, the AI-use statement and, for a file rendered from a release candidate, its release statement are titled back matter.
  • Body: sections nested by heading level, paragraphs, bulleted, numbered and checklist items, quotations, tables with their numbered captions, figures with their labels, captions and image files, equations as TeX, code and footnotes. A number bound to evidence is printed as its value.
  • References: one entry per cited work, built from the work's own record — authors, year, title, journal or publisher, volume, issue, pages, DOI, PMID and PMCID — with the reference exactly as your citation style prints it beside it. Every in-text citation links to its entry and reads as it does in the Word file.
  • What it refuses: tracked changes. JATS has no markup for a pending insertion or deletion, so a JATS export of the tracked-changes view is refused with TRACKED_CHANGES_NOT_SUPPORTED rather than written with the changes dropped; export all changes accepted or the original instead, or Word for a file that shows them.
  • What it leaves out: comments, authorship colours and page layout — page breaks, rules, a contents block — which belong to the editor, not the article. A picture placed inside a line of text is not exported in any format; the manifest names each one.

Validity is measured, not assumed. Each manuscript.xml is checked against the JATS 1.3 Journal Publishing DTD, which is kept on the server, with xmllint when the server has it. The result is recorded with the export — in manifest.json, in the export dialog, in the Release panel, in a journal package's checklist and in the answer your connected agent gets — as one of three outcomes: valid (the validator read this exact file and reported no error); not valid (with the validator's messages — the file is still produced, so you can see what is wrong, and should not be sent as JATS until it validates); or not run (no validator is installed on this server). A file is never called valid JATS when validation did not run.

A JATS file rendered from a release candidate carries the candidate's release statement and is recorded like every checked file. One exported from the live document is an UNCHECKED working copy and says so at the top of the XML, at the end of it and in its manifest.

Page view and layout

The editor lays the manuscript out as real pages: paper size (A4, US Letter, US Legal, A5, B5), orientation, margins, a header band and a footer band with page numbers, exactly as the export will print them. The Layout tab holds paper size, orientation, margin presets (Normal, Narrow, Moderate, Wide) or custom margins, header and footer text (use {page}, {pages}, {title} and {date}), the page-number position, line numbers (continuous or restarting on each page) and the switch between Page view and Web layout (one continuous column, handy on small screens). The status bar shows the current page; click it to jump to a page.

A hard page break (Insert → Page break) ends the page at that point in both the editor and the export.

The ribbon

On large screens the editor shows a ribbon. Home holds the clipboard, styles, fonts, formatting, paragraph controls, find and comments. Insert holds tables (a size picker or a custom size), figures from a file or an address, equations, quotes from PDFs, footnotes, links, captions, page breaks, dividers and code. Table appears while the caret is inside a table. Layout holds paper size, orientation, margins, headers and footers and line numbers. References holds citations, the bibliography and the citation style. Review holds the tracking mode, accept and reject (at the cursor or for the whole document), show-markup filters for insertions, deletions and formatting, comments, statistics, checks and versions. View holds the panels, focus mode, zoom, fit width, typewriter scrolling, page setup and the shortcut list. Phones and tablets keep the compact toolbar.

There is one more ribbon tab, Assistant, holding the AI writing actions — and it appears only where an administrator has switched the platform's own AI on. In the default mode you ask your connected agent for the same edits and they arrive as tracked changes.

Right-click anywhere in the text: the menu offers exactly what applies to the selection, the paragraph, the table, the figure or the tracked change under the pointer, including Accept and Reject. Its Assistant section — rephrase and the evidence checks — is there only where platform AI is switched on.

Proofing: grammar, spelling and the writing standard

The Review tab's Proofing group has three underlines, each switchable. Lints mark the writing standard — filler and model-typical wording, overstated conclusions, factual sentences with no citation — from the same checks the release gate runs. Grammar underlines grammar, spelling and style through LanguageTool running on this platform's own server (red for spelling, blue for grammar, grey for style): click an underline for the suggestions and pick one, or ignore it. The browser's own spell-check stays on as well. Statistics opens word counts and readability.

Next to the Grammar button, the language tells the checker what it is reading. *Detect* suits a text in one language; when a manuscript mixes languages or a short passage is misread, choose the language (English UK or US, Arabic, French, German, Spanish and others). The choice is remembered on this device and the open text is checked again at once.

Grammar checks are sent to this platform's API and nowhere else; the text is not stored. Citations, bound numbers and footnote marks are not words and are never flagged.

Figures and pictures

  • Insert from a file, by drag-and-drop or paste, or from a web address; every figure gets a numbered caption and alt text for accessibility and exports.
  • Select a figure for the picture tools: align left, centre or right; wrap text on either side (the picture floats and the text flows beside it); size presets (a third, half, three quarters, full width) or drag the side handles; rotate in 90° steps; crop by dragging the edge handles, with Reset crop; replace the image while keeping caption and settings; delete.
  • Figures are numbered automatically in reading order, and cross-references in captions follow when you move them.

Inline pictures and the Images panel

  • Insert → Inline places a small picture in the line of text (a symbol, a logo, a tiny chart) from a file or a web address; select it for size presets, drag-resize, alt text, aspect lock, or Convert to figure when it deserves a caption.
  • The Images panel (Insert → Images, or the Images tab on the left) lists every figure and inline picture with a thumbnail, caption and alt text; jump to one, fix alt text in place, or remove it. Pictures without alt text are counted so accessibility and journal requirements are met before submission.

Tables

  • Border presets (grid, horizontal rules, three-line journal style, none), banded rows, header row and header column, density (compact, normal, relaxed), width as a percentage of the text column and placement left, centre or right.
  • Cell shading, horizontal and vertical alignment; per-column alignment and number formats (integer, one to three decimals, percent, scientific) from Table properties.
  • Rows and columns, merge and split cells, split a table before a row, merge with the table below, convert selected text to a table and a table back to text, sort by a column, distribute columns evenly, add a totals row.
  • Captions with automatic numbering and a description read by screen readers and carried into exports.

Table of contents and abbreviations

  • References → Contents inserts a table of contents that follows the headings as you write: levels 1–3 by default, each entry a link to its heading, page numbers in page view. It exports as a plain list.
  • References → Abbreviations opens the abbreviations panel: every abbreviation with its expansion, number of uses and issues — used before it is defined, never defined, defined twice, two different expansions, defined but never used. Jump to the definition or the first use, Define at first use to move the expansion where it belongs, Ignore tokens that are not abbreviations, Copy list or Insert list as a table. Definitions are recognised as “Full term (ABBR)” or “ABBR (full term)” when the letters fit; the reference list is skipped.

Table painter and ruler

  • Table → Painter copies the current cell's shading and alignment; every cell you click afterwards receives them, until Esc or a second click on Painter.
  • Table → Ruler shows a ruler above the table with each column's width in millimetres and its share of the table; drag a marker to resize the two columns around it. Widths are stored with the table and respected by exports.

4.2 Citations and styles

Cite works from the project, switch styles without retyping, and let the reference list render itself.

Inserting a citation

Ctrl/Cmd + Shift + C or /cite opens the picker: search the project's references, look a paper up by DOI or title, or pick from evidence quotations you have already extracted. A citation may point at several works and carry a page or note.

Styles

Vancouver, AMA, APA 7, NEJM, Lancet, BMJ, Nature and Chicago; numeric styles renumber automatically and collapse ranges. Switching style changes every label and the reference list at once.

Integrity

A citation cannot be typed; it always references a record. When a project holds the same paper twice (the same DOI or PubMed ID, or the same title, year and first author), the References tab says so; merging keeps one record and the citations of either print as one reference with one number — nothing is rewritten and the merge can be undone. A work that is later retracted, superseded, corrected or placed under an expression of concern is flagged in a banner at the top of the editor and in the checks: a retraction or a supersession is an error that blocks a checked release, a concern or a correction is a warning. Numbers bound to evidence from that work stop binding, and any release candidate built on it goes stale. Files already exported or sent to a journal are never changed for you — you are told which version is affected and what to do. Removing a work from the project leaves its citations marked unresolved for you to fix.

Recovering citations from imported documents

When you import a Word file with numbered references, the reference list is parsed and numeric markers in the text are re-linked to works matched by DOI, PMID or title. Unmatched ones stay as text with a warning.

Inserting citations

Press ⌘⇧C (or References → Citation) to open the picker. It lists the project's references, everything else in your library and, on demand, a live search of PubMed, OpenAlex and Crossref; a DOI, PubMed ID or PMC ID pasted into the box is fetched and added in one step. Select several works for a grouped citation, give each one page numbers, a prefix ("see also") or a suffix, and choose the parenthetical or narrative form. Your PDF highlights can be inserted as quoted passages with their citation, and a reference that exists nowhere yet can be typed by hand.

Keyboard: ↑↓ move, space selects, Enter inserts, double-click cites one work immediately.

Editing a citation

Click any citation in the text. The popover shows what it cites with a link to the DOI, lets you change page numbers, prefix and suffix, remove one work or the whole citation, switch between (Smith 2020) and Smith (2020), jump to the entry in the reference list, or add and replace works through the picker. The reference list renumbers itself; nothing is stored in the text but the works.

Style and in-text format

References → Style opens the style chooser: every installed CSL style (Vancouver, APA, AMA, Nature and more) with a live preview. Numeric styles can print as [1], (1) or superscripts and collapse runs into ranges (1–3); author–date styles list references alphabetically and unnumbered. Changing the style re-renders every citation and the reference list at once, in the editor and in exports.

4.3 Reporting checklists

Attest your manuscript item by item against CONSORT, STROBE, PRISMA, CARE, SPIRIT, STARD, TRIPOD, ARRIVE, CHEERS and their extensions, and export the completed checklist for the journal.

Start a checklist

  • Project → Reporting checklist: choose the manuscript and the guideline that matches the study design. Thirty-five checklists are grouped by family — randomised trials (CONSORT 2010 and its AI, pilot, cluster, pragmatic and non-inferiority extensions, CENT, SPIRIT, SPIRIT-AI), observational and diagnostic (STROBE, STROBE-MR, STARD, TRIPOD, TRIPOD-AI, PROBAST-AI), reviews (PRISMA 2020, PRISMA-S, ScR, LSR, NMA, IPD, MOOSE, SANRA), case reports (CARE, CARE abstract, PROCESS), qualitative and implementation (COREQ, SRQR, GRAMMS, StaRI, SQUIRE), economics, animals and guidelines (CHEERS 2022, ARRIVE 2.0, RIGHT).
  • A manuscript can carry several checklists (a base guideline plus an extension); each appears as a tab with its completion.

Attest each item

  • For every item choose Reported, Partly reported, Not reported or N/A, and record where it is (section, page, line) and any note. Changes save as you type.
  • The progress bar counts addressed items over applicable ones (N/A items are excluded); the checklist is complete when nothing remains Not reported. Filter to Still open to work through what is left, or mark the remaining items as reported once you have checked them.

Export and the automated checks

  • Export CSV produces the completed table (item, status, location, notes) to attach to the submission.
  • Automated checks — required sections, journal limits, house style, citation integrity — run in the editor's Checks panel; the checklist is the human attestation that journals ask for alongside them.

4.4 Reporting and house-style checks

Run reporting checklists (CONSORT, PRISMA, STROBE and more), journal limits and writing checks before anyone else sees the draft.

Running checks

Open the Checks panel, or ask your connected agent to run them (run_checks). Results are grouped by severity; each item links to the paragraph it concerns and many offer a one-click fix as a tracked change. The checks themselves are deterministic — no model is involved — so they work in every mode.

What is checked

  • Reporting guideline items for the manuscript type (CONSORT, PRISMA 2020, STROBE, CARE, SPIRIT, ARRIVE, STARD, TRIPOD).
  • Journal limits from the front matter: words per section, abstract length, figure and table counts, keyword count, reference count.
  • Front matter completeness: authors with roles, disclosures, ethics and data statements, structured abstract sections.
  • House style: abbreviations defined at first use, consistent units and spelling variant, numbers at sentence start, p-value formatting, table and figure numbering and references to them.
  • Citation and evidence integrity: unresolved citations; citations to retracted, superseded, corrected or questioned works; numbers whose evidence came from such a work; numbers without a verified source.

Export gate

Errors block export; warnings do not. Either can be overridden with a reason that is recorded in the manuscript's audit log.

4.5 Journal finder

Journals ranked from where similar papers appear, where your cited works were published and how topics overlap — with the facts that show whether a journal is transparent, and how well the ranking did when it was tested on published papers.

Recommend for my manuscript

  • Project → Journal finder. Pick a manuscript (title and keywords are filled in) or type a title and a few keywords, optionally restrict to open access, a maximum article-processing charge or similar papers from a given year, then Recommend journals.
  • The finder searches OpenAlex for similar papers and looks at where they were published — about 150 from the title and keywords, or, when you give an abstract of at least 15 content words, a second search driven by the abstract, with the 50 papers closest to it deciding. Each journal's match (0–100) adds up: where similar papers appear (45), where the works already in your project were published (25), topic overlap with your keywords (20) and impact (10). The reasons are spelled out on every card.
  • Cards show articles indexed, citations, h-index, two-year mean citedness, APC, access (subscription, open access, DOAJ-listed) and topics, with links to the journal site and its OpenAlex record.

Transparency signals

  • Instead of a blacklist, the finder lists verifiable facts: open access without a DOAJ listing, very small output, an unusually high APC, a low h-index for the journal's size, a generic all-encompassing title, no ISSN. DOAJ listing, a high h-index and a large output count in a journal's favour.
  • The caution level (none, minor, check carefully, serious) summarises those facts. It is a prompt to look — at the editorial board, the peer-review statement and your library's guidance — never a verdict.

Compare and add as target

  • Compare up to four journals side by side (publisher, articles, citations, h-index, citedness, access, APC, cautions).
  • Add as target creates a submission target under Submissions with the journal's name, ISSN and homepage; add the journal's limits and reference style there, and the pre-flight and the formatter use them.

How accurate is it?

The finder's accuracy is measured, not assumed. An administrator draws published papers at random from OpenAlex — research articles in English with an abstract, published in a journal over recent years, drawn with a fixed seed so the same set can be drawn again — and the finder is run on each of them, given its title, its abstract and, where OpenAlex lists them, its keywords. The run records, paper by paper, which of them came with keywords.

  • The paper being tested is held out: it is kept out of the papers the finder compares against, by its OpenAlex id, its DOI and its exact title (which also catches a second OpenAlex record of the same paper). Otherwise the paper would point at its own journal and the test would only measure that.
  • For each paper the test records where the journal that actually published it appears in the finder's ranking — matched by OpenAlex id, or by ISSN when OpenAlex keeps the journal under another record — or that it does not appear at all.
  • A run reports how often that journal is first, in the first five and in the first ten (the top-10 rate), each with a 95% confidence interval, and the median rank where it was ranked. The rates count only the papers that could be evaluated: a paper for which OpenAlex did not answer is counted as not evaluated — never as a miss and never as a hit — and drawn papers that could not be a test at all (no journal, or no abstract long enough for the finder to read) are counted as omitted.
  • The finder shows the latest completed run that evaluated at least 30 papers, in one sentence — “In the last test on N published papers the true journal was in the top 10 for X% of them (date)” — with how many more could not be evaluated and what the finder was given. A smaller run is an administrator's trial: it stays in Administration → Evals and is never shown here, so the figure does not swing with an experiment on a handful of papers. Until a run of that size has completed, the finder says its accuracy has not been measured yet; a run that was stopped or interrupted is never shown.
  • Administration → Evals keeps every run's history: the papers, the seed and the OpenAlex filter they were drawn with, the finder version each run measured, and which run the finder page shows.

The figure is about the ranking: how often the journal that published a paper is among the first ten the finder lists for it. It says nothing about what an editor will decide about your manuscript, and it was measured with each paper's title and abstract (and keywords where OpenAlex lists them) — a ranking from a title alone may do worse.

What the finder does not have

  • Scopus and Web of Science journal lists: neither is available to the product, so the finder never says whether a journal is indexed there. Check the journal's own site or your library.
  • Time to first decision and time to publication: no lawful public source covers them for journals in general, so the finder shows neither.
  • What it does show comes from OpenAlex (papers, journals, citations, fees where recorded), DOAJ and the NLM Catalog (MEDLINE).

4.6 Internal peer review

Review rounds on a manuscript before submission: invite colleagues, collect structured reviews with recommendations and scores, keep identities blind if you wish, decide, and export the report.

Rounds

  • Project → Peer review: pick the manuscript and open a round with a type — team (everyone sees who wrote what), blind (reviewer names hidden from everyone but the round's manager and project leads) or open (all reviews visible to all) — an optional deadline and a brief for reviewers.
  • Invite reviewers from the project's members; they are notified and accept or decline. Add people to the project from the Team tab first (the reviewer title is made for this: comment and screen without editing).

Writing a review

  • Reviewers choose a recommendation (accept, minor revision, major revision, reject), score originality, methodology, clarity and significance from 1 to 5, and write a summary, strengths, weaknesses, suggestions and confidential comments to the editor.
  • Inline comments left in the manuscript during the round are part of the review — reviewers open the manuscript from the round.
  • Leads are notified when a review arrives; in blind rounds the notification does not name the reviewer.

Decision and report

  • The round's manager closes it with a decision and a summary for authors and reviewers; rounds can be reopened.
  • Report exports the round as Markdown — decision, every review with its scores and sections (confidential comments only for managers) — to file with the submission or share with co-authors. Journal reviewer comments after submission live under Submissions.

4.7 Journal submissions

Track a manuscript from target journal to decision: pre-flight, packages, author agreements and reviewer responses.

Creating a submission

  1. Submissions tab → New submission; pick the manuscript version and the target journal.
  2. Pre-flight checks the manuscript against the journal's limits and your checklist and lists what still needs work.
  3. Collect author agreements: each co-author confirms authorship, conflicts and the submitted version from their notifications.
  4. Build the package: manuscript in the journal's format, cover letter, figures, tables, supplements and the AI-use statement, zipped. JATS XML of the clean manuscript (manuscript.xml with its figures) is added when the journal's profile lists JATS among its files, or when you tick Include JATS XML; the checklist says why it is or is not there and whether it passed DTD validation. Write the cover letter yourself or ask your connected agent for a draft; the “Draft with the assistant” button beside it needs platform AI switched on.

Status and history

Draft → ready → submitted → under review → revision requested → resubmitted → accepted, rejected or withdrawn. Every transition is logged with who and when; the team is notified.

Revisions

Import the decision letter: it is kept verbatim with its checksum, and each comment is decomposed into the separate things it asks for. You record what you decided about each one; the response letter is then composed from those records against the release candidate you are sending, and travels in the package beside the reviewers' own letter. See Responding to reviewers.

4.8 Responding to reviewers

A point-by-point response composed from records: it can claim a change only where the release candidate you are sending actually carries it.

From letter to atomic requests

Import the decision letter and it is stored exactly as it arrived, with its sha256; nothing ever rewrites it. Each reviewer's comments are separated on their headings, and each comment is decomposed into the separate things it asks for — “add a confidence interval” and “report the absolute risk difference” are two requests, not one — with a kind (a text change, an analysis to run, data or code to supply, something to state, a citation, or a request you may reasonably decline) and the section or table it points at. Every request's text is a span of the reviewer's own words. Your connected agent can re-kind or re-split a comment; it cannot put its own words in one.

Deciding, not drafting

  • There is no box to write the reply in. You record a disposition for each request — changed, will change, disagree, out of scope or deferred — with your reason, and (for “changed”) the tracked changes that made it. The sentences the reviewer reads are composed from that.
  • A change addressed to a reviewer request is made with the REVIEWER_REVISION mode, which records the request ids it answers. That record is what lets the letter attribute the change to the point; an edit made at the same time but naming nothing cannot answer it.
  • Only a signed-in person may record a disposition: answering a reviewer is a scientific decision about their objection. A connected agent may propose one, and the proposal is stored beside your record, never over it.
  • A disagreement is a legitimate answer and needs no change — but it must state the scientific reason, and a blank reason is refused. A silent refusal is not an option the product offers.

Composing the letter

  • Compile the response against the release candidate you intend to send. Each point that claims a change is checked against that candidate's frozen document: the change set must be accepted, must name this request, and must be found in the document — by the edited block's hash, by the inserted text, or as an unresolved tracked insertion.
  • Locations — “page 4, lines 12–18” — are read from the candidate's own PDF render, never typed. Where a passage could not be matched in the text layer the letter says so instead of printing a number, and the numbers are of Hikma's render, not of the journal's own pagination.
  • An analysis is claimed only where the run succeeded, and the run is named.
  • A claim the candidate does not carry is printed unresolved with the reason and fails the check RESPONSE_UNBACKED, which refuses the submission package. There is no waiver for it: it is about a sentence addressed to somebody else.

Four separate channels

  • Manuscript — the clean current text. It never says that a passage was added or clarified for a reviewer; the REVISION_LANGUAGE check fails a release that does.
  • Tracked view — the same document rendered with the changes shown. A projection, not a second document.
  • Response letter — composed from records, as above.
  • Revision ledger — the change sets with the mode each declared and the requests each names.
  • Editorial annotation — comment threads in the manuscript, which are part of neither the manuscript nor the letter.

Export

The compiled letter as DOCX or Markdown. The submission package carries both, plus the reviewers' letter verbatim, and the checklist names every file with its sha256 and says how many points the response answers with a change.

5. Collaboration

Messages, notifications, profiles and following.

5.1 Messages

Direct and group conversations with requests, live delivery, reactions, attachments, mute, block and report.

Starting a conversation

Messages → New, or the Message button on a profile. People you share a project with can be messaged directly; anyone else receives a request they accept or decline. The recipient's contact preference (anyone, people they follow, project connections, no one) is enforced.

In a conversation

  • Live delivery, typing indicators and seen-by receipts.
  • Replies, reactions, editing and deleting your own messages (deleted for everyone or hidden for you).
  • Attachments up to 25 MB, ten per message, with image previews.
  • Groups with names, avatars, admins and a system log of who joined or left.
  • Mute for an hour, a day, a week or until you turn it back on; pin and archive conversations.

Safety

Block someone to stop all contact silently; report a conversation or a message with a reason and administrators review it. Away-only notifications mean you are not pinged for messages you are already reading.

5.2 Notifications

One inbox for everything that concerns you, delivered live, by e-mail instantly or as a digest, with per-project mutes.

What you are told about

AreaEvents
ManuscriptsProposals left by a co-author or an AI agent, comments and mentions, restores, submission status, author agreements, exports
ProjectsDiscussion messages and mentions, screening conflicts, finished analyses, members joining, role changes, meetings, tasks, deadlines, full-text results
MessagesDirect or group messages while you are away, requests, being added to a group
CommunityNew followers, likes, comments, replies, reposts and mentions in the feed
SystemAnnouncements and support replies

Channels

Every event can reach you in the app and by e-mail, switched separately per event under Settings → Notifications. E-mail can be instant or bundled into a daily or weekly digest at the hour you choose, in your time zone.

Live

The bell updates without reloading, a toast appears for anything outside the page you are on, and the notifications page groups everything by day with filters by area and unread only.

Muting a project

From the project menu, from the bell on any notification, or under Settings → Notifications → Muted projects. Muted projects create no notifications at all until unmuted.

Reading and clearing

  • Opening a notification marks it read; Mark all read clears the badge at once and Clear read removes them from the list.
  • Dismiss single items you do not need again. Older items load on demand; the list keeps everything until you clear it.
  • The bell's dropdown shows the latest few; the Notifications page is the full inbox with filters by area and unread only.

5.3 Profiles and following

Public researcher pages with publications, metrics and badges; follow colleagues and control who sees what.

Your public page

Name, credentials, position, institution, location, bio, research interests, methods, languages and links, plus impact (publications, citations, h-index, projects, collaborators) and badges computed from your activity (collaborator, prolific researcher, highly cited, rising scholar, mentor, peer reviewer, community builder, ORCID linked).

Who can see it

SettingAudience
Anyone with the linkPublic, including people without an account
Signed-in researchersAny account holder
Only you and people you work withMembers of projects you share

Open-to flags (collaboration, mentoring, reviewing) are shown as chips so the right people reach out.

Following

Follow a researcher to be notified when they publish or join projects you can see, and to allow them to message you when your contact preference is “people I follow”. Followers and following lists are visible on the profile.

Publications and metrics

Claim papers by DOI or PubMed ID, by name or ORCID iD, or import them from ORCID. Citations, the h-index and the i10-index are computed from the claimed papers’ own citation counts (the higher of what OpenAlex and Crossref report for each paper); an author-level total is used only when no paper has a count yet, and every card says which. Refresh the counts from Settings → Publications.

A metric shown as “—” means the citation counts have not been fetched yet, not that the number is zero.

5.4 Teams

Research groups across projects: members and roles, a forum with accepted answers, live channels, shared files and linked projects.

Kinds of team

TypeWho can joinListed under Find teams
PublicAnyone, immediatelyYes
InstitutionalPeople with the creator's e-mail domainYes, to that institution
PrivateBy request (admins approve) or invite linkYes
SecretInvite link onlyNo

Roles

The owner can do everything including transferring ownership, archiving and deleting. Admins approve requests, manage members and invite links, open channels, pin topics and post announcements. Members post, reply, vote, upload files and link projects they belong to.

Forum

  • Categories: question, discussion, announcement (admins), resource.
  • Questions can have an accepted answer, chosen by the asker or an admin; answered questions are marked in the list and the answerer is notified.
  • Upvote topics and replies; pin what everyone should read; search titles, bodies and tags; filter to unanswered questions.

Channels

Channels are team conversations on the messaging engine: live delivery, replies, reactions, attachments and read receipts. Every member is in every channel; leaving the team leaves its channels. Open them from the team page or from Messages.

Files and projects

Upload protocols, forms, slides and data (200 MB per file) into folders; files are served with short-lived links to members only. Link the projects the team runs so newcomers find the work; members of a project open it, others see its public record and can ask to join.

Project ideas and volunteering

  1. Team → Ideas → Share an idea: title, category, what you have and what is missing, and the roles you need volunteers for (statistician, second reviewer, clinician…). Every member is notified and the team's activity records it.
  2. Every role is a tag with its own Volunteer button; any number of members may volunteer for the same role (or name another role), each with a line about what they bring. The author is notified.
  3. The author reviews volunteers per role in arrival order — each carries its position and a time tag, so first come, first served is visible — and presses Accept or Decline (each person is told), then Start the project when the team is there: a private project is created with the author as principal investigator and every accepted volunteer as co-investigator. Once the idea is a project the same role buttons read Join as …: accepted people are added to the project, and one removed later leaves it.
  4. Ideas shared with a team also appear on the idea board, but only to that team's members; a private team's pitch is never public.

Team → Overview shows the ideas still looking for volunteers, so a newcomer sees at once where help is wanted.

Joining and inviting

  1. Find teams lists public, institutional and private teams; join or request in one click.
  2. Admins accept or decline requests under Members; requesters are notified either way.
  3. Invite links (Members → Invite links) let anyone join directly, including secret teams; links can expire, be limited to a number of uses, and be revoked.

5.5 Idea board

Pitch a study, method, tool or dataset; gather collaborators with the skills you lack; turn the idea into a project when the team is there.

Posting an idea

  1. Idea board → Post an idea: title, category, a description that says what you have and what is missing, roles and skills needed, tags.
  2. Colleagues upvote, comment and apply with a role and a note.
  3. Accept the applications you want; accepted people appear on the idea as the team.

Ideas posted here are visible to everyone signed in. To pitch inside a team only — with members volunteering for named roles — use the team's Ideas tab (see Teams); such ideas show a Team badge on the board and only that team's members see them.

Turning it into a project

When the team is there, Turn into a project creates a private project with you as principal investigator and every accepted applicant as a co-investigator, and marks the idea as converted with a link to the project. From there the usual workspace applies: tasks, sources, evidence, manuscripts.

Status

Open (looking for collaborators) · in progress · converted · closed. Only open ideas accept applications; you can withdraw a pending application at any time.

Board etiquette

  • One idea per post; split if you have more than one.
  • Say what you already have (data, access, funding, a draft) and what is missing — applicants decide from that.
  • Close or convert the idea once the team is complete so people stop applying; withdraw applications you no longer mean.
  • Keep confidential details for the project you create afterwards; ideas are visible to every signed-in member.

5.6 Mentorship

Find researchers open to mentoring, ask with concrete goals, and keep goals, meetings and resources in one place.

Who can mentor

Anyone who switches on Open to mentoring under Settings → Research. Mentors are ranked for you by how well their interests and methods overlap with the topics you enter, with a small preference for mid-career and senior researchers.

Asking

  1. Mentorship → Find a mentor: enter the topics and methods you want help with, pick a mentor and Ask.
  2. Write concrete, time-bound goals and how often you would like to meet; your time zone is shared for scheduling.
  3. The mentor accepts or declines; you are notified either way and can ask someone else.

Working together

Active mentorships carry shared notes typed as goal, meeting, resource or note. Goals can be ticked off; both sides see everything. Either side can end the mentorship; the notes stay readable.

5.7 Conferences

A shared directory of meetings and deadlines with reminders, and your abstract submissions tracked from draft to presented.

The directory

Anyone can add a conference: name, place or online, dates, abstract deadline, website and topics. Filter upcoming or past, search by name, city or topic. The person who added it (or an administrator) can edit it.

Interest and reminders

Mark a conference as interesting to see who else is going and to be notified about a week before abstracts close.

Your submissions

  1. On a conference page, New abstract: title, type (poster, oral, workshop, symposium) and the abstract text.
  2. Move it through draft → submitted → accepted or rejected → presented as things happen.
  3. Attach the poster or slides when they are ready; files are private to you.

5.8 PDF requests

Ask an author who claimed a paper for its full text when no open-access copy exists; authors answer with the PDF or a note.

Asking

On a publication page without an open-access copy, Ask an author for the PDF sends a request to a researcher who claimed the paper. Add a short note; the author is notified and sees the request under PDF requests → Asked of me.

Answering

Authors upload the PDF (shared with the requester alone, through a short-lived link) or decline with a note, for example when publisher terms do not allow sharing. Requesters can cancel pending requests.

Sharing responsibly

Share only what your licence or publisher agreement allows — author manuscripts are usually fine, publisher PDFs often are not. The platform records who shared what with whom.

5.9 Home feed and Discover

The home feed gathers what you follow and what happens in your projects; Discover finds people, papers, projects and teams across the platform and the literature.

Posting, threads, media and sharing

The feed works like the networks you know, kept for research. A post can carry text, up to four images or short videos (paste or attach), and a paper card. Like it, comment on it — comments have their own likes and replies — or open the thread to reply with a full post of your own. Share with a repost or a quote, save it for later, or copy the link. Every card has a Report action; reports and automatic flags reach the moderators.

  • Tabs — For you is every public post; Following adds your projects' activity and the publications the people you follow claim; Saved holds what you bookmarked. Announcements stay pinned above all three.
  • @mentions — type @ and pick a colleague; their name becomes a link to their profile and they are notified. Renaming an account never breaks the link, because the person, not the text, is stored.
  • #hashtags — write #hypertension anywhere in a post and it becomes a link to every post carrying that tag. Trending this week beside the feed lists the tags used most in the last seven days.
  • Save — the bookmark icon keeps a post in the Saved tab; saving is private to you.
  • Edit and delete — rewrite or remove your own posts and comments at any time; an edited item is marked “edited” with the time. Moderators can delete anything that breaks the guidelines.
  • Media — click an image to open it full screen (arrow keys move between images, Escape closes); videos play in place with full controls. Add alternative text in the composer so people using a screen reader know what you shared.
  • Who to follow — suggestions beside the feed from your institution, your research interests and the people your own network follows.
  • Followers-only posts stay with your followers. Media is stored privately on the platform and served only to signed-in members.
  • A post held for review by an automatic rule is visible to you with a “held for review” note until a moderator restores it; editing a post runs the same rules again.

Keep posts to findings, questions, papers and calls for collaboration; the community guidelines and the moderators handle the rest.

What appears in the home feed

  • Posts from people you follow — a thought, a question, a paper with your commentary (attach one from your library or by DOI).
  • Publications your colleagues claim on their profiles.
  • Activity in your projects: manuscripts created, members joining, submissions moving, analyses finishing.
  • Administrator announcements, pinned above the stream.
  • Switch between For you, Following and Saved at the top, or filter any of them to one hashtag. The feed loads more as you scroll, and a Load more button does the same by hand.

Posting and engaging

  • New post: text with @mentions and #hashtags, up to four images or short videos with alternative text, and an optional attached paper.
  • Like, comment (threaded), repost, quote and save. The author is notified of likes, comments, replies and shares, and anyone you mention is notified too — repeats are folded into one notification (“Ada and 3 others liked your post”).
  • Edit or delete your own posts and comments; edits are marked and go through the moderation rules again.
  • Administrators can hide or delete reported content and the author is told why.

Discover

  • Researchers — search by name, institution, interests or methods; filter by open to collaboration, mentoring or reviewing; follow from the card.
  • Publications — search the literature across the backbone sources (OpenAlex, PubMed, Semantic Scholar, Crossref) without opening a project; save any result to your library or open its publication page to recommend it, discuss it or ask an author for the PDF.
  • Projects — public projects looking for members, with a join request.
  • Teams — public teams to join or request to join.
  • Claim publications — type your name as it appears on papers, pick your author record (OpenAlex), tick the papers that are yours and claim them in one go; they join your profile, portfolio, metrics and CV. Link your ORCID under Settings → Publications to import everything at once.
  • Global search (the header box or ⌘K) reaches the same people, papers, projects, manuscripts, notes, teams and courses you can see — see the Search guide.

Following and public pages

Follow a researcher from their profile or a Discover card: their posts and claimed publications join your feed and they are notified, with no access to anything private. Your own public page shows what you allow under Settings → Profile (see Public profile).

Feed versus notifications

The feed is for ambient awareness. Anything addressed to you — mentions, comments on your work, requests, assignments, deadlines — arrives in Notifications with a badge, so nothing that needs action can be missed in the stream.

6. Academy

Courses, resources, quizzes, certificates and the blog.

6.1 The Academy

Courses built from lessons, resources and quizzes; standalone videos, articles, PDFs and links; a blog; certificates you can verify.

What is in it

The Academy is the platform's learning space. Anyone can publish; everything published is visible to every signed-in member.

ContentWhat it isWhere
CourseAn ordered set of modules — text lessons, resources (video, article, PDF, link), quizzes, or a Lesson that combines text, a video or resource and a quiz in one module — with enrolment, progress and a certificateAcademy → Courses
ResourceA video, article, PDF, document or link that stands alone or sits inside a courseAcademy → Resources
Blog postAn article with tags, likes and commentsAcademy → Blog

Learning

  1. Open a course and press Enrol — courses are free.
  2. Work through the modules in order; mark lessons and resources complete, pass quizzes.
  3. When every required module is done the course is complete and your certificate is issued automatically.
  • Continue where you left off from the Academy home or My learning.
  • Keep private notes on any module; they stay with the module.
  • Save any course, resource or post to find it later under Saved.
  • Ask a question on a course — the author is notified, and their reply is marked as the accepted answer.
  • Rate and review a course once you are enrolled.

Certificates

Each certificate carries a twelve-character verification code. Anyone can check a code at /certificate/<code> without signing in; the page shows the learner, course, author and issue date. Print or save the certificate as PDF from the certificate page.

Optional modules do not block completion. Authors mark a module optional when it is supplementary.

6.2 Creating a course

Details, modules of three kinds, quizzes with a pass mark, publishing, and the learners report.

Details first

  1. Academy → Create → New course. Give it a title, description, difficulty, language, topic, tags, learning objectives and prerequisites.
  2. Add modules: a text lesson (rich text with headings, lists, tables and links), a resource (pick any published resource, or one of your drafts), or a quiz.
  3. Reorder with the arrows, mark modules optional where appropriate, then Publish. A course needs at least one module to publish.

Quizzes

A quiz module holds multiple-choice and true/false questions, each with exactly one correct answer and an optional explanation shown after answering. Learners complete the module by reaching the pass mark (default 70%) and can retry.

Put a short quiz after each block of lessons rather than one long one at the end; completion rates are far higher.

Publishing states

StateMeaning
Draft (private)Only you see it; you can preview every module as a learner.
PublishedListed in the catalogue, on your instructor page and in search.
ArchivedHidden from the catalogue; enrolled learners keep access and certificates stay valid.

Your learners

The edit page shows who enrolled, when, how far they got, their best quiz score and their last activity. Questions from learners reach you as notifications; your reply is marked as the accepted answer.

6.3 Resources and the blog

Publish standalone videos, articles, PDFs and links, and write posts with tags, likes and comments.

Resources

  • Video — paste a YouTube or Vimeo link (embedded without cookies) or upload an MP4/WebM up to 200 MB; add a transcript to make it searchable.
  • Article — written in the rich editor.
  • PDF or document — uploaded; PDFs preview inline and can be downloaded.
  • Link — an external page.
  • Resources have drafts, topics, tags, a duration, comments and bookmarks. A resource used by a course module is shown inside the course.

Blog posts

Write in the rich editor, add an excerpt and tags, save drafts and publish when ready. Reading time is computed. Readers can like, comment, save and follow tags to related posts.

7. Account and security

Profile, sign-in methods, two-factor, sessions, privacy.

7.1 Plans, credits and billing

What your plan includes, credits and codes, changing or cancelling a plan, invoices — and why connecting your own AI agent costs nothing here.

When billing is switched off

Billing is switched off: everyone has every feature. The pricing page shows the planned plans without prices, and codes you redeem now are kept on your account for launch.

Your own AI agent is not metered

By default the platform runs no models of its own: you connect Claude Code, Codex, Cursor, Claude Desktop, ChatGPT or another MCP client, your provider bills its reasoning as it does today, and everything the agent does through Hikma — searching the literature, fetching full texts, reading, extracting evidence, writing manuscripts, screening, pooling, analysing data — is free and unmetered. No subscription and no card are needed for it; see Connect an AI agent (MCP) and the API.

How the platform's own AI is counted

This applies only where an administrator has switched the platform's own assistant on (Admin → AI & keys) and billing is live; while billing is off nothing here limits you. Each of the platform's model calls is metered at the provider's cost in US dollars and recorded against the run, the project and you. Your plan includes an allowance per period; credits cover use beyond it; when both are used up the platform's assistant pauses until the period resets, you add credits, or you upgrade. Unattended runs are limited per period on some plans.

Settings → Subscription shows the allowance bar, credits, unattended runs used and the reset date. On a bring-your-own-agent deployment the allowance simply never moves.

Changing plans

  1. Settings → Subscription → pick a plan → checkout opens in Stripe.
  2. Manage billing opens the Stripe portal for cards, addresses and receipts.
  3. Cancel at period end keeps everything until the paid period is over; Resume renewal undoes it.

Invoices appear on the same page with links to the hosted invoice and PDF.

Codes and credits

  • Credit codes add credit for the platform's own AI at once.
  • Free-days codes unlock a plan for a number of days without a card.
  • Discount codes apply at your next checkout — enter the code before choosing the plan.
  • Credit packs (one-time top-ups) are sold when billing is on.

7.2 Settings overview

Where each preference lives: profile, research, publications, CV, notifications, security, integrations, AI memory and activity.

The sections

SectionContents
ProfilePhoto, identity, affiliation, links, who can message you, profile visibility
ResearchInterests, methods, discipline, career stage, open-to flags, ORCID linking
PublicationsClaimed papers, ORCID import, citation metrics
CV & biosketchEducation, positions, grants, awards, presentations; Academic, NIH and NSF exports
NotificationsPer-event in-app and e-mail switches, digests, muted projects
SecurityPassword, two-factor, passkeys, sessions, linked sign-ins, e-mail address, account deletion
IntegrationsAPI keys, connected AI clients and the MCP recipes for each of them
Assistant memoryYour AI preferences and what your AI is allowed to remember about you
ActivityYour footprint, storage used and safety counters

Storage allowance

Each person has a storage allowance — 5 GB by default, set by the administrator — covering everything they upload: project and team files, datasets, message and post attachments, and figures. Settings → Activity shows how much of it is used; deleting a file gives the space back. Full texts the platform fetches for you do not count.

Display name and e-mail

Your display name is shown everywhere; changing it in Profile updates projects and messages. The e-mail address is changed under Security and, once verified, requires approval from the current address.

7.3 Account security

Strong passwords, two-factor codes with backup codes, passkeys, institution single sign-on, sessions you can see and revoke.

Passwords

At least ten characters; common passwords, your name and your e-mail are rejected both in the browser and on the server. Changing your password signs out every other device. Reset links are valid for one hour; reset also signs everything else out.

Two-factor authentication

  1. Settings → Security → Turn on two-factor; confirm your password.
  2. Scan the QR code with any authenticator app (or type the key) and enter the six-digit code.
  3. Save the backup codes; each works once if you lose the app.
  4. From then on, password sign-ins ask for a code. Tick “Trust this device” to skip it for 30 days on that device.

Two-factor protects password sign-ins. Passkeys are already phishing-resistant and do not ask for a code.

Passkeys

Face ID, Touch ID, Windows Hello or a hardware key. Add one per device (or a synced passkey from your password manager), then use “Continue with a passkey” on the sign-in page. Remove lost devices from the list.

Institution sign-in and linked accounts

If your administrator registered your institution's identity provider, “Continue with your institution” signs you in with your work address. Google and ORCID can be linked under Security for sign-in; ORCID does not share e-mail addresses, so create the account first, then link.

Devices and sessions

Every signed-in device is listed with browser, system, IP address and last activity. Sign out one, or all others. Sessions expire after 30 days of inactivity.

Protection against guessing

Sign-in, sign-up, reset and code endpoints are rate-limited per network address, and too many wrong codes lock two-factor for a while. Suspicious activity is visible in your sessions list.

7.4 Privacy and your data

What is stored, who can see it, how models are used, and how to export or delete.

What we store

  • Account: name, e-mail, sign-in methods, sessions.
  • Profile: what you enter, publications you claim, metrics fetched from OpenAlex.
  • Projects: everything members put in them, with an activity log.
  • AI conversations: the tool calls a connected agent made and their results, so work can be audited — plus prompts and costs per run where the platform's own AI is switched on. What you type into your own client stays with your provider.
  • Messages and notifications, with read state.

Language models

By default the platform sends nothing to a language model. When you connect your own agent, that agent reads what it asks for through the MCP server — sources, evidence, manuscript passages — and sends it to your provider under the terms and the plan you already have with them; Hikma is not a party to that and never sees your prompts.

Where an administrator has switched the platform's own AI on, the platform makes the calls: the text it sends is limited to what a task needs, the provider and keys are the administrator's, each run shows the model used, and providers are contractually barred from training on the data.

Your controls

  • Profile visibility and contact preference (Settings → Profile).
  • AI memory: see, confirm and forget everything your AI is allowed to remember, or switch memory off.
  • Muting, blocking and reporting in messages and projects.
  • Delete account (Settings → Security): removes profile, memberships, messages and keys; shared project content stays with the project attributed to a former member.

7.5 Your public profile

What other researchers see: identity, badges, publications with search and sort, impact analytics, projects and groups — and exactly how each metric is computed.

The five tabs

  • Overview — name, credentials, position and career stage, institution and location, ORCID, bio, interests, keywords, methods, languages and external links; headline metrics and the most recent papers; projects you share with the viewer.
  • Publications — every claimed paper with search (title, author, journal, DOI), year and type filters and sorting by newest, most cited, title or recently claimed; each row links to the DOI and can be saved to your library.
  • Impact — the metrics below, papers per year, citations per year, the citation distribution, top journals, frequent co-authors (co-authors with an account here link to their profile) and the most cited papers.
  • Portfolio — the same claimed works as a one-page, print-ready research portfolio with a copyable public link.
  • Projects & groups — the projects and groups this researcher belongs to that you are allowed to see.

How each metric is computed

Every citation number on the profile, the Portfolio tab, Settings → Publications and the CV comes from one shared definition, so the same person can never show two different totals.

MetricDefinition
PublicationsPapers claimed on this profile.
Total citationsThe sum of each claimed paper's own citation count. Each paper keeps the highest count reported by OpenAlex or Crossref. Where not one paper has a count yet, the author-level total from a linked ORCID record is shown instead and labelled as such.
Citations per paperTotal citations ÷ publications, to one decimal.
h-indexThe largest h for which h claimed papers have at least h citations each.
i10-indexClaimed papers with ten or more citations.
Active since / active yearsThe earliest publication year, and the span from the earliest to the latest.

Each card says where its number came from — "from 56 papers" or "author-level count from OpenAlex". When citation counts have not been fetched yet the h-index and i10-index show "—", never 0, with a link to refresh them; when only some papers have counts the profile says so and treats the totals as a lower bound.

Refresh citation counts from Settings → Publications after claiming new papers. The refresh also corrects the journal of papers that were listed under a repository such as figshare, bioRxiv or SSRN.

Who sees what

Profile visibility (public, members only, private) is set in Settings → Profile, as is who may message you. Private projects and groups only appear to people who share them. Followers and following lists open from the counters under the header.

7.6 Publications and CV

Claim your papers, keep metrics current, and generate an academic CV, NIH biosketch or NSF biographical sketch.

Claiming publications

Paste DOIs or PubMed IDs, one per line, or import everything from ORCID. Metadata comes from the backbone sources and is de-duplicated into canonical works, the same records used in project libraries. Each paper's citation count is fetched as it is claimed, so your metrics are right straight away.

Finding yourself by name: type your name as it appears on papers — or your ORCID iD, which matches exactly one author record — and, if you share a name with other researchers, your institution. Each candidate shows its affiliations, how many works and citations it has, its h-index and i10-index, what it works on, and why it was ranked where it was.

Metrics and refreshing them

Total citations, the h-index and the i10-index are computed from your claimed papers' own citation counts — the same definition the public profile, the Portfolio tab and the CV use. Each paper keeps the highest count reported by OpenAlex or Crossref, so a refresh can never quietly lose citations.

  1. Claim your papers (ORCID import, by name, by title or by identifier).
  2. Press Refresh citation counts to re-read every claimed paper from OpenAlex and Crossref; the card shows when this last ran and how many papers have a count.
  3. Press Recompute metrics to rewrite the stored h-index, i10-index and totals from those counts. Linking your ORCID also brings in the citations-per-year series from your author record.

A refresh also repairs venues: a paper listed under a repository (figshare, bioRxiv, medRxiv, SSRN, Zenodo…) is re-labelled with the journal that published it, and a paper that only exists in a repository says so explicitly rather than presenting the repository as a journal.

If the h-index reads "—", the citation counts have not been fetched yet. Refresh them; the profile never shows 0 for a number it does not know.

CV and biosketches

  1. Settings → CV & biosketch: enter education, positions, grants, awards and presentations once.
  2. Add a personal statement and synergistic activities for funders.
  3. Pick Academic CV, NIH Biosketch or NSF Biographical Sketch; preview updates as you save.
  4. Download as DOCX (editable in Word) or print to PDF.

Publications, identity and metrics are pulled from your profile, so the CV never drifts from the rest of your record.

Claiming from Discover and the portfolio

  • Discover → Claim publications finds your author records by name, ORCID iD or institution (also alternative name forms and affiliations) and lets you claim papers in bulk; already-claimed papers are marked. Papers without a DOI can be added by identifier under Settings → Publications.
  • Your profile's Portfolio tab turns the claimed works into a one-page research portfolio — identity, publications, citations, h-index, i10-index, active since, publications by year, where the work appears, frequent co-authors, most-cited papers and the full list — with Print / save as PDF and a copyable public link. It reads from the same claimed works as your CV.

8. AI, integrations and administration

Connecting your own AI agent, MCP clients, API keys and administration.

8.1 Desktop app for Windows, macOS and Linux

Analyse data on your own computer with the same statistics as Data, open large CSV, Excel and Parquet files out of core, send results to a project, and write manuscripts in the built-in editor with offline autosave.

What the desktop app is for

The desktop app runs Data's analyses on your computer: your file never leaves the machine unless you choose to send a result to a project. It opens CSV, TSV, Excel, SPSS (.sav, .zsav), Stata (.dta), R (.rds) and Parquet files, reads files larger than memory through an embedded database, and runs descriptives, tests, regression, survival and mixed models with APA tables, charts and the syntax to reproduce them.

  • Open data… picks a file; a spreadsheet asks which sheet. The Variable View shows labels, types and the case weight, as in SPSS. Save data as… writes the active cases and every variable, computed ones included, as a .sav, .dta, .rds or .csv file with the labels and value labels set here (SPSS also keeps the missing values and the weight; R gets factors for coded variables).
  • Analyze lists every procedure; Guide suggests one from a plain-language question about your data.
  • Write opens your manuscripts from this platform in a window of the app, signed in as on the web. Edits made without a connection are kept on the device and sent when it returns.
  • hikmaresearch.org sends an output to a project as a note or a dataset output, with a personal API key you create under Settings → API keys.

Computed variables and the Syntax tab

From version 0.1.5 the app has Data's command language. Compute… makes a new variable from a formula over the others (arithmetic, comparisons, AND/OR/NOT and the usual functions); the Syntax tab runs a whole plan as one script — ANALYZE, TITLE, SPLIT FILE, WEIGHT, FILTER BY a formula, COMPUTE and TRANSFORM (recode, z-scores, ranks, bins, lags, imputation and more) — checked line by line as you type and run through the same engine as the Analyze menu. Every output has Open in Syntax, and Add from outputs turns your session into a script.

  • Computed and transformed variables are definitions kept in the project file, recomputed in order whenever an analysis reads them; the data file is never changed, and the Data View shows the file as it is.
  • A z-score or a rank is always taken over every case, not over a sample: when a computed variable or a formula filter is involved, the app loads the needed columns whole (up to a million rows) before filtering or sampling.
  • The same script runs unchanged on hikmaresearch.org, where R and Python program blocks also run; here the checker says on the line what cannot run on this computer (dummy coding, dataset-creating case operations, R/Python).

Installing

  1. Download the installer for your system from the Downloads page (hikmaresearch.org/downloads, also linked under Settings → Integrations): Windows .msi or .exe, macOS .dmg, Linux .AppImage, .deb or .rpm.
  2. First launch of an unsigned build — Windows: SmartScreen says "Windows protected your PC"; choose *More info → Run anyway*. macOS: do not double-click the first time — right-click (Control-click) *Hikma Research* in Applications and choose *Open*, then *Open* again; if macOS says the app is "damaged", run xattr -dr com.apple.quarantine "/Applications/Hikma Research.app" in Terminal once. Linux: .deb/.rpm install normally; make the AppImage executable (chmod +x) before running it. Code signing and notarisation are on the roadmap and need the project's own certificates.
  3. Start the app. Nothing needs a sign-in until you send a result to a project or open Write.

The app talks to this platform over HTTPS only — the same addresses your browser uses. It never holds server credentials beyond the personal API key you give it, and it never opens an SSH connection.

Large files

Files up to the size of your disk open through DuckDB: the grid pages through rows, filters and sorts run in the database, and descriptives on wide tables sample up to two thousand values per column for the preview while the analysis itself uses every row. Saving a project keeps the file path, the variable dictionary and the recent outputs.

FormatDesktop appBrowser preview
CSV / TSVYes, any sizeYes, in memory
Excel (.xlsx)Yes, by sheetNo
ParquetYesNo
SPSS .sav / .zsavYes, with labels, value labels, missing codes and the weightYes, in memory
Stata .dta (Stata 13+)Yes, with labels and value labelsYes, in memory
R .rds (a data frame)Yes, with factor levels, labels and haven value labelsYes, in memory

What leaves the computer

  • Nothing, by default. Analyses run locally and outputs are written to your disk.
  • What you send with hikmaresearch.org → Send: the chosen output (summary, tables as text, syntax) to the project you pick.
  • What the Write window does: it is the web application, so manuscripts live on the platform as they do in a browser, with a local copy kept for offline editing.

8.2 Your AI agent

AI reaches this platform through an agent you already have — Claude Code, Codex, Cursor, Claude Desktop, ChatGPT or any MCP client. It connects to your projects, reads the sources and evidence and writes into your manuscripts as tracked changes. What an administrator adds by switching the platform's own assistant on is set out at the end.

How AI works here

By default this platform runs no models of its own. It holds the research — the works, the full texts, the evidence units, the screening decisions, the datasets, the manuscripts — and exposes all of it to your agent through one MCP server. The reasoning happens in the client you already use and pay for; the data, the rules and the record stay here.

  • You choose the model. Claude Code, Claude Desktop, claude.ai, ChatGPT, Codex, Cursor, VS Code, Windsurf, Zed, Cline — any client that speaks the Model Context Protocol.
  • You pay your own provider. Hikma meters nothing for this; a free account and either an OAuth sign-in from the client or a personal key are enough.
  • It acts as you. Exactly your permissions in exactly your projects, every call recorded under your account on the project's Activity tab.

Bring your own agent is this platform's default. An administrator can switch the platform's own assistant on as well — see “When the platform's own AI is switched on” at the foot of this page. That adds a second way of working beside your agent; it does not change anything in the sections above.

Connecting an agent

  1. Open a project → Your AI (the AI section of the project bar). A project with no conversations opens straight on the connection guide; otherwise press Connect your AI.
  2. Pick your client. The panel writes out the exact command or configuration file with this deployment's server address already filled in.
  3. Choose how it signs in: OAuth through the client (your browser opens and you approve it) or a personal key created in the same step and shown once.
  4. Choose which tools it gets — everything, or the smaller Literature & writing, Systematic review or Data analysis packs for clients that struggle with long tool lists.
  5. Run the command. The panel waits and confirms as soon as the first call arrives.

The same panel is under Settings → Integrations, and one connection covers every project you belong to: you name the project in what you ask, not in the configuration. Per-client recipes — Claude Code, Claude Desktop, claude.ai, ChatGPT, Codex, Cursor, Windsurf, VS Code, Zed, Cline and stdio-only clients — are on Connect an AI agent (MCP) and the API.

What it can do, and what it cannot

Connected, an agent gets 180 tools. All of them are things the platform does without a model of its own:

  • Literature — search the backbone sources, plan and run a search strategy, snowball, add and remove works, list what still has no full text.
  • Reading — acquire and fetch full texts, read a source page by page or as structured XML, search inside a source's text.
  • Evidence and review — read the evidence table, set the protocol, record screening decisions, save risk-of-bias and GRADE judgements, insert the review tables.
  • Writing — create and read manuscripts (whole blocks, never shortened; each block with a stable id and a content hash, and with per-sentence provenance on request), propose paragraphs, patch a sentence or replace text bound to the exact text it read, insert tables, update the front matter, run the checks.
  • Data — register and describe datasets, run catalogue analyses and sandboxed Python or R, insert results and figures into a manuscript.
  • The project — list projects and the workspace, keep a plan, write notes, post messages into the conversation, store a memory proposal.

What it cannot do matters as much:

  • It cannot cite a work that is not in the project library. Citations are objects bound to works; text an agent types is not a citation.
  • It cannot type an unsourced number into a manuscript. Numbers arrive as bound numbers pointing at an evidence unit, a page or an analysis result.
  • It cannot overwrite your text. Everything it writes is a tracked change.
  • It cannot decide a screening call, a risk-of-bias judgement or a GRADE rating. What it records is a vote or a suggestion until a reviewer confirms it.
  • It cannot read a paper the platform could not fetch. list_missing_full_texts names those works with the reason; upload the PDF on the Sources page and ask again.
  • It cannot read a document you uploaded until you say it may. A file you upload is recorded as your licensed copy: Hikma reads, stores, OCRs, indexes and quotes it for your research, and does not hand its text to your agent. See “What your agent may read” below.

Its own abilities are welcome: the server tells the agent to use its own web search and page fetching for discovery and to register whatever it finds with add_works or fetch_article, so every citation still resolves to a project work with provenance.

What your agent may read

Every document in a project carries a rights class, and the class decides — per document, per person, per destination, on every call — whether its text may leave for your agent. Retrieval, reading on screen, server-side processing, storage, OCR, indexing, quotation, delivery to an external AI, export and redistribution are separate rights: a permission for one is never a permission for another.

The document isYour agent may read it
Open access with a reuse licence (CC-BY, CC0, public domain)Yes, in full, immediately. Export and redistribution are permitted too, except under a NoDerivatives licence, which permits quotation but not rendering the text into an exported file.
Open access, free to read, no reuse licenceYes, in full. It is not exported or redistributed as a file.
A copy you uploadedYes, at once. Your upload is the statement that you are entitled to work on the document here, so there is no attestation step on any upload path — yours in the browser, or your agent's over MCP under your key — and the uploader is recorded as provenance. The other members of the project read it with their own agents too. It is still not exported and not handed to another project without a rights decision.
A copy whose rights are not recordedA bounded excerpt only — the smaller of 15% of the document and 2,000 characters in total, counted across every tool and every page. Reading it in smaller pieces does not enlarge it.
Licensed through a publisher agreementNo. Publisher text-mining terms prohibit uploading their content to an AI system; Hikma's own extraction still reads it and returns anchored values and quotations.

A refused read is never silent. Your agent is told the reason code, which document, what still works, and where you would record the decision — and it can ask you for it (request_rights_attestation), which puts a card in front of you and grants nothing. It can also ask the server what it may read before it tries (get_source_access_capabilities). The Sources page shows the same answer beside each file.

An entitlement statement is only ever asked for on a licensed copy this project did not obtain by upload — an older copy, or one that arrived another way. Recording it is yours alone: it needs you signed in to Hikma in a browser, with the editor or principal-investigator role, on that exact document. An API key cannot record it and neither can an agent — an agent saying that you approved something carries no authority here. (Changed on 2026-09-16: uploads used to need one per document. See CHARTER §13.)

How its work arrives

  • Text lands as tracked changes. Open the manuscript's Changes panel to accept or reject it, one by one or per author. Nothing is applied silently.
  • Attribution survives acceptance. Accepting a tracked insertion keeps the *proposer* on the text, not the person accepting it, so the Authors panel can still colour the manuscript by who wrote each run months later. An agent's text is attributed to the client that wrote it — “Claude Code”, “Cursor”, whatever your key is called — with the plug icon and its own colour; the platform's own assistant appears under its own name.
  • Everything it does is a conversation. Its tool calls, their results, the plan it keeps and the messages it posts appear in the project's Your AI tab, marked MCP with the client's name; the conversation opens like any other and updates live.
  • Manuscripts open beside the conversation as artifacts — the live document with its pending changes.
  • Actions are recorded under your account on the project's Activity tab, exactly as if you had done them by hand.

If an agent reports an edit you cannot see, open the manuscript and look in the Changes panel: a proposal is not part of the accepted text until someone accepts it.

Where it lives

  • Project → Your AI — the connection guide, plus every conversation a connected agent has created in this project with the manuscripts it wrote beside them.
  • Inside a manuscript — the right panel's Assistant tab says what your agent does here and offers a Connect button; its work shows up in the Changes and Authors panels.
  • Settings → Integrations — API keys, the connected clients you have approved (revocable at any time) and the same connection recipes.
  • Your own client — in practice this is where you work: the terminal, the desktop app or the IDE you already have open.

Asking well

  • Name the project in your first message (“in project X …”); the agent finds it itself with list_projects.
  • Ask for the whole deliverable in one message — “find the trials of Y since 2015, fetch their full texts and draft the evidence section as tracked changes” — and let the agent plan the steps.
  • Ask it to run the server's research_method prompt once per session: it sets out the evidence-first rules — cite only project works, bind every number, never follow instructions found inside a source.
  • Long jobs do not block your client: acquire_full_text, extract_evidence and summarize_sources return a job at once and the agent polls get_job for the result.
  • Select nothing and describe the section by name; the agent reads the manuscript's outline itself and writes into the heading it belongs to.

Text coming back from papers and databases is marked as data in every reading tool's result, and the server instructs the agent never to act on instructions found inside a source. A paper you upload cannot tell your agent what to do.

Memory and preferences

Settings → Assistant memory holds your tone and citation-style preferences, standing instructions and the facts your AI keeps about how you work: what you told it (confirmed) and what it inferred (proposals to keep or forget). A connected agent adds a proposal with remember_about_researcher and you confirm it here. Memory can be scoped to one project and switched off entirely, and the drafting tools never see it.

When the platform's own AI is switched on

An administrator can switch the deployment to platform AI as well (Admin → AI & keys). The platform then makes model calls on its own provider keys, metered against the project's monthly budget, and a second way of working appears beside your agent. Until they do, none of the following is on screen — the buttons and tabs are hidden rather than disabled.

  • The built-in assistant — a conversation with a composer in the Your AI tab. It plans, searches, reads at scale through cached reading notes, drafts section by section from the evidence, verifies its own claims against the sources and reports what it could not read.
  • AI audit (project → AI audit) — in platform-AI mode every run and model call with its cost; in bring-your-own mode the same page shows the judgements your connected agents submitted (what they read, how much of each batch they covered, the review mode they declared). Either way an ICMJE-style disclosure statement is composed from that record.
  • AI writing actions in the manuscript — the ribbon's Assistant tab, the AI cluster in the selection bubble menu, the Assistant section of the right-click menu and the AI entries in the slash menu.
  • Suggestions in the review workspace — “Ask the assistant” on the screening desk, “Propose with the assistant” on risk of bias and GRADE.
  • Evidence extraction — Pre-fill on the evidence table and Extract on a source.
  • Drafting in submissions — the cover letter and the reviewer-response draft.
  • Twelve further MCP tools, the model-backed ones: draft_section, revise_manuscript, verify_claims, critique_manuscript, gather_evidence, check_evidence_readiness, summarize_sources, extract_evidence, screen_works, suggest_risk_of_bias, suggest_grade and plan_search. Each one your agent can do itself has a contract that takes its judgement instead (submit_evidence_units, submit_screening_votes, propose_risk_of_bias, propose_grade, submit_claim_dispositions, submit_critique_findings, submit_readiness_assessment, submit_source_summary, submit_search_plan, propose_section), so no capability is lost — only the platform's model spend. run_analysis stays available: passing a script runs it in the sandbox, which needs no model.

A built-in conversation is a run with a budget set by the project (defaults from the administrator). It works in rounds until it delivers its report or the budget is spent, continues on the server if you close the page, and can be stopped at the current step; steps, rounds and cost sit under each answer. Every model call is metered — but only the platform's own calls. Your connected agent is never metered by Hikma, in either mode.

Everything in the sections above still holds whichever engine writes: text arrives as tracked changes, citations are objects, numbers are bound to their evidence, and screening and grading stay suggestions until a reviewer confirms them.

You can tell which mode a deployment is in without asking anyone: in bring-your-own mode the AI section of the project bar holds Your AI and the AI audit of what your agents submitted but no Workflows, the Your AI tab opens on the connection guide with no composer, and your MCP client shows server instructions that say the client is the only reasoning engine and list the submission contracts.

8.3 AI audit and disclosure

What was written with AI help and how to disclose it: the conversations a connected agent leaves in every project, the judgements it submitted, and — where an administrator has switched platform AI on — a costed audit tab. Both compose an ICMJE-style statement.

The record

Whichever engine does the thinking, the project keeps the trail:

  • Conversations. Project → Your AI lists every conversation, including those a connected MCP client created — marked MCP with the client's name — with its tool calls, their results and the plan it kept. A conversation closes after 30 minutes without a call and stays readable.
  • Activity. Every action an agent took is on the project's Activity tab under your account, next to the same actions taken by hand.
  • Authorship. The manuscript's Authors panel colours the text by who wrote it and keeps that attribution after a tracked change is accepted, so the share written with AI help is visible in the document itself rather than only in a log.

The AI audit tab

Project → AI audit shows whichever record exists. Where an administrator has switched platform AI on (Admin → AI & keys) it reports the platform's own model spend. In the default bring-your-own-agent mode there is no model spend to meter — your agent's calls are billed by your own provider and never reach the platform's ledger — and the tab reports the judgements your agents submitted instead.

  • Platform AI. Every run of the platform's assistant (interactive, unattended, workflow) with status, spend and a link to its transcript, and model use by task class and model with calls, failures, tokens and spend. The record comes from the ledger that meters every call; nothing is inferred.
  • Your own AI. Every judgement a connected client handed in through a submission contract — extracted evidence, screening votes, a risk-of-bias or GRADE proposal, claim dispositions, a critique, a readiness assessment, a reading note, a drafted section — with the client that submitted it, what it says it read (source versions, pages, works, the manuscript content hash), how much of the batch it covered, and, for a review, the mode it declared (self-review, external second opinion, isolated review, human expert review).
  • A review mode is what the client said, not something Hikma measured: a second token, client, session or model proves attribution, not independence. The page says so, and never claims more.

Disclosure statement

Where platform AI is on, a statement is composed from that record — the tool, the models, what they were used for, the period and the number of calls — following the ICMJE recommendation on AI-assisted technology. Copy it into the acknowledgements or methods and adjust the wording to the journal's template.

Where you brought your own agent, the export names the client — Claude Code, Cursor, whatever your key is called — and how many proposals it made and how many you accepted, all from the record. The one thing the platform cannot see is which model your client ran, so the statement leaves that blank for you to complete; everything else is filled in. Where an administrator has switched the platform's own AI on, the models are named from the ledger and nothing is left blank.

8.4 Connect an AI agent (MCP) and the API

Step-by-step: connect Claude Code, Claude Desktop and claude.ai, OpenAI Codex, Cursor, Windsurf, VS Code, Zed, Cline, ChatGPT or any MCP client to your projects over MCP — with OAuth or an API key — and what the agent can do once connected.

What you get

  • Hikma Research exposes one MCP server (Model Context Protocol, streamable HTTP). Any MCP-capable agent — a coding assistant in your terminal, a desktop assistant, an IDE — reaches your projects with it: search the literature and add works, fetch full texts and read them page by page or as structured XML, read the evidence table, plan and record screening, save risk-of-bias and GRADE judgements, run meta-analyses and analysis plans, write and edit manuscript text as tracked changes, run the checks, list projects and tasks. This is how AI works on this platform by default: your client does the reasoning, Hikma holds the research.
  • The agent acts as you: exactly your permissions in exactly your projects, every action logged under your account on the project’s Activity tab. The same guard-rails apply — it cannot cite a work outside the project library or type an unsourced number into a manuscript, and its edits arrive as tracked changes for you to accept.
  • Your server address is shown under Settings → Integrations (it ends in /mcp). On a personal machine it is http://localhost:4100/mcp; on a deployment it is your API's public HTTPS address plus /mcp.

Your subscription, not ours

Connecting your own agent needs no Hikma subscription and no card: a free account and a personal key (or an OAuth sign-in from the client) are enough. Your agent's reasoning is billed by its own provider exactly as it is today; everything it does through Hikma — searching nine bibliographic sources, fetching full texts, reading and extracting evidence, writing manuscripts as tracked changes, screening, pooling, analysing data — is free.

  • Bring-your-own-agent mode (the default). The platform spends no model keys, so the tools that would have done the thinking are not offered — and each names the contract that takes your agent's judgement instead. Your agent reads sources with read_source_page, read_source_xml and search_source_text, reasons itself, and hands in what it concluded: submit_evidence_units (the server checks each quote against the page and decides verified / partial / unverified / contradicted / not_reported), propose_section (the server binds the numbers and reports every digit that bound to nothing), submit_screening_votes, propose_risk_of_bias, propose_grade, submit_claim_dispositions, submit_critique_findings, submit_readiness_assessment, submit_source_summary and submit_search_plan. Nothing is metered by Hikma.
  • Platform AI mode. An administrator may additionally switch on Hikma's own assistant (Admin → AI & keys). Then tools such as draft_section, gather_evidence, verify_claims, critique_manuscript and screen_works also appear over MCP; they run on the platform's keys and count against the project's monthly budget, never against you personally while billing is off. The submission contracts stay available in both modes — an agent that read the papers itself should be able to say so.
  • Either way the agent sees exactly your projects with exactly your role, and every call is recorded as a conversation in the project's Your AI tab.

You can tell which mode a deployment runs from the server's instructions (your client shows them — in bring-your-own mode they name each absent tool and the contract that replaces it) or from a project's Your AI tab: the tab opens on the connection guide and has no composer, and the project bar's AI section holds Your AI and the AI audit of what your agents submitted.

The easy way: from the Your AI tab

Open a project → Your AI. A project without conversations shows the Connect panel first; afterwards the Connect your AI button above the conversations opens the same steps. Choose your client, which tools it gets, and how it signs in; the panel shows the exact command or configuration with the server address (and your new key, if you chose one) filled in, then confirms the connection when the first call arrives. What the agent does then appears as a conversation in that project, and the manuscripts it writes open beside the chat.

  • Toolsets. Everything is the default (192 tools when the administrator has switched platform AI on; 180 in bring-your-own mode, where the model-backed ones are replaced by the submission contracts). Literature & writing (about half the schema size), Systematic review and Data analysis are smaller packs for clients that struggle with many tools; a key remembers the pack you chose, and any client can ask for a pack with the header X-Hikma-Tools: writing | review | data | all (Claude Code: --header).
  • Online. The address shown is your deployment's public API address (https://api.hikmaresearch.org/mcp on the production platform). Hosted clients — Claude Desktop, claude.ai, ChatGPT — sign in through OAuth against it; local development servers are only reachable from clients on the same machine.

Two ways to authenticate

  • OAuth (recommended for people): add the server by URL; the client opens your browser, you sign in and approve the client, and it receives tokens bound to your account. Revoke a client any time under Settings → Integrations → Connected clients.
  • API key (for scripts, servers and clients without OAuth): Settings → Integrations → New key (choose an expiry of 30–365 days), then send it as an Authorization: Bearer header. Keys are limited to 600 requests per minute and can be revoked at once; treat them like passwords and never commit them.
  • Clients that only speak stdio (no HTTP) reach the server through the mcp-remote bridge: npx -y mcp-remote <server-url> — add --header "Authorization: Bearer <KEY>" for a key, or let it run the OAuth flow.

Claude Code (terminal)

# OAuth: add, then run /mcp inside Claude Code and choose Authenticate claude mcp add --transport http hikma https://api.hikmaresearch.org/mcp # API key instead of OAuth claude mcp add --transport http hikma https://api.hikmaresearch.org/mcp --header "Authorization: Bearer <KEY>" # check claude mcp list

bash

Then ask in plain language: “List my projects”, “Search PubMed and OpenAlex for prophylactic amiodarone after cardiac surgery and add the ten most relevant trials to project X”, “Extract the evidence table from the included works”, “Draft the Background from the evidence and propose it as tracked changes”. The research_method prompt on the server explains the evidence-first rules to the agent; run it once per session.

Claude Desktop and claude.ai

  • Settings → Connectors → Add custom connector → paste your server address (a public HTTPS address is required; localhost is not reachable from the hosted app) → Add → sign in and approve when the browser opens.
  • Enable the connector in a chat with the tools icon. Everything you ask runs under your account; disable or remove the connector from the same screen.
  • Claude Desktop also accepts a JSON configuration with the mcp-remote bridge for local development: see the generic section below.

OpenAI Codex

# ~/.codex/config.toml [mcp_servers.hikma] url = "https://api.hikmaresearch.org/mcp" # either let Codex run the OAuth flow (codex mcp login hikma), or send an API key: # http_headers = { Authorization = "Bearer <KEY>" }

toml

Recent Codex CLI versions also accept the command line: codex mcp add hikma --url https://api.hikmaresearch.org/mcp. Check codex mcp list, then ask Codex to call the Hikma tools like any other MCP server.

Cursor, Windsurf, VS Code, Zed, Cline and others

// Cursor — .cursor/mcp.json (project) or ~/.cursor/mcp.json (global) { "mcpServers": { "hikma": { "url": "https://api.hikmaresearch.org/mcp" } } } // with an API key: { "url": "...", "headers": { "Authorization": "Bearer <KEY>" } } // Windsurf — ~/.codeium/windsurf/mcp_config.json { "mcpServers": { "hikma": { "serverUrl": "https://api.hikmaresearch.org/mcp" } } } // VS Code (Copilot agent mode) — .vscode/mcp.json { "servers": { "hikma": { "type": "http", "url": "https://api.hikmaresearch.org/mcp" } } } // Zed — settings.json (Zed reaches HTTP servers through the mcp-remote bridge) { "context_servers": { "hikma": { "source": "custom", "command": "npx", "args": ["-y", "mcp-remote", "https://api.hikmaresearch.org/mcp"] } } } // Cline / Roo Code / Continue / Gemini CLI and any stdio-only client { "mcpServers": { "hikma": { "command": "npx", "args": ["-y", "mcp-remote", "https://api.hikmaresearch.org/mcp"] } } }

json

Each client shows the Hikma tools in its tools list after the first successful call. The exact file names and keys follow each client's current release; the shapes above are the ones in use at the time of writing — when in doubt, the client's own MCP documentation wins, and the server address never changes.

ChatGPT

In ChatGPT, Settings → Connectors (developer mode must be enabled by your workspace) → Create → paste the server address → connect with OAuth. A public HTTPS address is required.

What the agent can call

AreaTools
Projectslist_projects, get_workspace, update_plan, request_user_input, remember_about_researcher
Literaturesearch_literature, plan_search, run_search_plan, snowball, add_works, remove_works, list_works, get_works, acquire_full_text, fetch_article, list_missing_full_texts
Readinglist_sources, read_source_page, read_source_xml, search_source_text, summarize_sources, submit_source_summary, get_source_notes, gather_evidence, extract_evidence, submit_evidence_units, get_evidence_table, get_job, get_source_access_capabilities, request_rights_attestation
Writingcreate_manuscript, read_manuscript, draft_section, propose_section, propose_paragraphs, replace_text, insert_table, update_front_matter, run_checks, check_evidence_readiness, submit_readiness_assessment, verify_claims, submit_claim_dispositions, critique_manuscript, submit_critique_findings, check_review_article
Reviewset_review_protocol, screen_works, submit_screening_votes, record_screening_decision, get_screening_queue, get_screening_summary, list_included_studies, risk of bias and GRADE tools (set_* for reviewers, propose_* for an agent), insert_review_tables
Analysisregister_dataset, describe_dataset, create_analysis_plan, run_analysis, run_dataset_analysis, run_dataset_code, transform_dataset, run_meta_analysis, pool_meta_analysis, insert_analysis_output
Publishingrecommend_journals, search_journals, reporting checklists, peer-review rounds, notes
Collaboration and the review interior (M3 parity)list_tasks, create_task, update_task, list_deadlines, list_meetings, list_project_files, list_discussions, post_discussion, list_project_activity, list_annotations, create_annotation, update_note, delete_note, search_workspace, list_memory, forget_about_researcher, import_references, get_search_log, update_work_notes, reparse_source, request_missing_paper, create_project, get_review_protocol, export_review_protocol, set_search_strategy, log_search_run, export_search_appendix, list_screening_conflicts, set_full_text_decision, record_grey_literature, set_prisma_counts, render_prisma_diagram, get_review_audit, get_analysis_figure, edit_dataset, run_sample_size, run_randomisation, list_forms, get_form_responses, export_reporting_checklist, run_submission_preflight, list_manuscript_versions, compare_manuscript_versions, list_change_sets, set_change_set_status, list_manuscript_comments, add_manuscript_comment, render_working_export
Method profileslist_method_profiles, check_method_profile, set_method_label

Tool names as listed by the server are authoritative (ask the agent to list them). Articles are fetched as structured XML whenever a source offers it: fetch_article takes a DOI, PubMed id, PMCID, arXiv id or URL, walks the open-access ladder (Europe PMC / PubMed Central, Europe PMC preprints, bioRxiv/medRxiv, PLOS, Frontiers, PeerJ, MDPI, Copernicus, then PDFs through GROBID), waits for parsing and returns the section map; read_source_xml serves the JATS or TEI markup in slices for agents that prefer it to page text.

Long-running tools do not block the client: acquire_full_text, extract_evidence and summarize_sources return a job at once, and over MCP so do draft_section, gather_evidence, verify_claims, critique_manuscript and check_evidence_readiness (their model calls can outlast a client's 60-second request limit). The agent polls get_job with waitSeconds up to 45 and reads the tool's result there; the outcome lands in the project like any other action.

Tool names as listed by the server are authoritative (ask the agent to list them). Long-running tools return a job that the agent polls; results land in the project like any other action.

Security

  • Every request is authenticated — an OAuth access token bound to your account and the approved client, or an API key (hashed at rest, shown once, 90-day expiry by default, revocable, last use visible under Settings → Integrations). Unauthenticated calls get 401; the whole API is rate-limited per address.
  • Authorisation is per project. Each tool checks your role in the project it touches (viewer, editor, owner); an agent using your key sees exactly what you see, and calls on projects you do not belong to are refused before anything is recorded.
  • Inputs are validated (typed schemas on every tool) and all database access is parameterised. Uploads are checked for type and size and stored content-addressed.
  • Prompt injection. Text coming back from documents and databases is marked as data in every reading tool's result, and the server's instructions tell the agent never to follow instructions found in sources. Messages an agent posts are rendered as plain text and Markdown, never as HTML.
  • Outbound fetches are guarded. When Hikma downloads a paper it only follows public http(s) addresses; private, loopback, link-local and cloud-metadata destinations are refused at every redirect hop, and DNS answers are checked.
  • Recording is scoped. MCP conversations are created only in projects you belong to, under your user, and inherit the project's budget; long tools run as jobs you can only read back yourself.
  • Online. Publish the API over HTTPS behind your domain; hosted clients (Claude Desktop, claude.ai, ChatGPT) need that public address. Keep the OpenAlex key and other provider keys in the admin key store — never in a client configuration.

Troubleshooting

  • 401 or “unauthorized” — the token expired or the client was revoked: in Claude Code run /mcp and authenticate again; in other clients remove and re-add the server, or create a fresh API key.
  • The hosted Claude or ChatGPT cannot reach localhost: use your deployment's public HTTPS address, or test with Claude Code, Codex or an IDE on the same machine.
  • 429 — an API key is capped at 600 requests per minute; unattended loops should pace themselves.
  • The agent proposes edits you cannot see — open the manuscript: proposals appear as tracked changes in the Changes panel; nothing is applied silently.
  • Ask the agent to run the research_method prompt first if it starts inventing citations; the rules stop it from doing so through the tools regardless.

8.5 Administration

For administrators: users, audit log, analytics, system health, flags, models and keys, evaluations, moderation and support — all stored in the database, no env files.

Who sees it

Accounts with the administrator role see Administration in the sidebar. The operator names the first administrators in the deployment's environment (HIKMA_ADMIN_EMAILS) — they are promoted the moment they sign up; further administrators are promoted under Users. Every administrative action is written to the audit log with the actor, the target and the details.

Users

  • Search by name or e-mail; filter by role and status (active, suspended, e-mail not verified).
  • Manage opens the account in tabs: Overview (counts, institution, role, verification and two-factor state, plan), Projects & work, Activity (audited actions), Sessions, and Plan & credits (subscription, monthly model spend, grant credits). Support actions: mark the e-mail verified when the message never arrived, reset two-factor for someone locked out, sign out everywhere, suspend with a reason and optional length. The list accepts ?q= in the URL so other pages can link straight to an account. Also shown: projects, manuscripts, library items, model spend, sessions with device and address, recent actions and reports against the person.
  • Change the role, sign the person out everywhere, or suspend with a reason and an optional duration. Suspension takes effect immediately and blocks sign-in until lifted or expired.

Audit log

Filter by actor, action, target type and date range; export the current filter as CSV (up to 5,000 rows). Actions are dotted names such as admin.user.suspend or manuscript.export.

Analytics

Sign-ups and active users per day for the chosen range, with totals for projects, manuscripts, library additions, course enrolments, posts, messages, model calls, spend and tokens, and open reports. The chart has a table view for screen readers and copying.

System

  • Health and latency of PostgreSQL (with database size), Valkey, object storage, GROBID and the collaboration server.
  • Slowest procedures: every API call is timed; the table shows calls, p50, p95 and maximum per procedure with slow (> 500 ms) and error counts since the last reset — the budget is p95 under 300 ms for reads and 500 ms for writes.
  • Every background queue with waiting, active, delayed, completed and failed counts; failed jobs list their reason and can be retried in one click.
  • The API process: uptime, memory, Node version, environment and version. Refreshes every 30 seconds.

AI mode, provider keys and the task-class registry

  • AI mode decides who pays for the thinking. *Bring your own agent* (the default): researchers connect Claude Code, Codex, Cursor, Claude Desktop, ChatGPT or any MCP client on their own subscription; the built-in assistant and every model-backed tool are hidden, and the stored keys serve only embeddings (semantic search) and your tests. *Platform AI as well*: the built-in assistant, AI writing actions, screening, extraction, verification and unattended runs run on the keys stored here, metered by the budgets at the bottom of the page. The switch applies within seconds to the web app and to every MCP connection.
  • Provider keys are grouped the way you shop for them — major vendors (Anthropic, OpenAI, Google, xAI, Mistral, Cohere), China (DeepSeek, Qwen, Z.AI GLM, Moonshot Kimi, MiniMax, Doubao, Baidu Qianfan, Tencent Hunyuan, StepFun, SiliconFlow, iFlytek Spark), aggregators and fast inference (OpenRouter, Groq, Together, Fireworks, Cerebras, SambaNova, NVIDIA NIM, Perplexity) and self-hosted (Ollama, any OpenAI-compatible endpoint such as vLLM or LM Studio). Each card links to the vendor's key page, presets the API address (editable for regional endpoints), and once a key is stored fetches the live model list, so the pickers below offer real ids. Keys are encrypted at rest and never shown again; Test sends a one-line request to the model you pick and reports latency and cost; Disable keeps a key but makes routing skip it; Remove deletes it.
  • Bibliographic source keys (NCBI, Semantic Scholar, OpenAlex, the Unpaywall contact e-mail) are optional: searches work through the public tiers without them, a key only raises the rate limit or unlocks a premium tier for everyone.
  • The task-class registry lists every model dispatch in the platform (search planning, screening, extraction, source notes, classification, verification judge, critique, assistant planning, drafting, revision, analysis code, embeddings) with its purpose. For each: provider and model, an ordered fallback chain (providers without a key are skipped), an on/off switch and a per-person daily cap. Changes apply within 15 seconds without a restart. The verification judge should not share a model family with drafting.
  • Budget defaults — per narrative review, per systematic review and per project per month — apply to the platform's own model calls only.

AI costs

The platform's own model spend. In bring-your-own-agent mode there is nothing to report here: researchers’ agents are billed by their own providers, and only embeddings and your provider tests reach the ledger.

  • Windows of 24 hours, 7, 30 or 90 days: spend, calls, failed calls, tokens (with the share served from the prompt cache), average and p95 latency, month-to-date spend against the monthly budget default and the alert threshold from Settings.
  • Spend and calls over time; spend by task class with a 30-day trend, by provider and model, top people (opens their account) and top projects; prompt-cache hit rate per provider; the twenty most recent failures with their error text.
  • Export every call in the window as CSV (up to 20,000 rows) for finance or audit.

Quality

Platform-AI deployments only — it reads the run ledger, which stays empty while researchers bring their own agents.

  • The last 50–500 finished assistant runs across every workflow: success rate, median cost and duration, tool calls per run, budget and round-limit stops, model failures and rate-limited runs, checks passed versus failed.
  • Breakdown by workflow and mode, the most frequent failure reasons, claim–evidence verification links for the last 30 days by relation and acceptance, and manuscript checks by identifier and severity.
  • Each run links to its transcript in the project's assistant. Read-only; use it to compare prompt or routing changes before and after.

Mentorship

  • Open or close the programme: while closed the directory hides and no new requests are accepted; active mentorships continue.
  • Listed mentors with interests, methods and workload (active, pending, completed); Unlist removes someone from the directory without touching the rest of their profile.
  • Every request with its state (looking for a mentor, awaiting the mentor's answer, active, declined, ended); End closes one on the programme's behalf with an audited reason.

Feature flags

Flags gate features at runtime by key; toggling takes effect within seconds and needs no deploy. Add a description so the next operator knows what a flag does.

Sign-in providers and e-mail

Google and ORCID client credentials, institution SSO providers (OIDC, by e-mail domain) and the outgoing SMTP server are platform settings under Administration → Settings. Until SMTP is configured, account e-mails are logged instead of sent and the settings pages say so.

Billing

Administration → Billing: switch billing on for launch, store the Stripe secret and webhook signing secret (encrypted), edit plans (price, assistant allowance, unattended runs, project and collaborator limits, features) and sync them to Stripe with one click, define credit packs, create named or bulk single-use codes (credits, free days, percentage or fixed discounts), grant credits to a person, and watch subscriptions and revenue. While billing is off, the whole platform is free and the pricing page shows plans without prices.

Evaluations

Evals runs the platform's own assistant against a fixed suite of research tasks and reports groundedness, citation validity and cost per task, so a model or routing change can be judged before it reaches researchers. It needs platform AI switched on and provider keys stored; it says nothing about the agents researchers connect themselves.

  • Journal finder accuracy needs no AI and no model key: draw published papers from OpenAlex (how many, which publication years, a seed), then run the finder on them. Each paper is held out of its own evidence, and the run records where the finder ranks the journal that published it. The finder page shows the latest completed run that evaluated at least 30 papers; a smaller run is a trial, kept in the history here and marked as not shown. See Journal finder → How accurate is it?
  • A run works in the background — closing the page does not stop it — and the page follows its progress: papers evaluated, papers that could not be evaluated, and the final top-1, top-5 and top-10 rates with their intervals and the median rank. Stop ends a run after the paper in hand; a stopped or interrupted run is kept in the history and never shown as a measurement.
  • A run asks OpenAlex about three times per paper. OpenAlex meters requests: store an OpenAlex key under AI & keys before running a large suite, or the run can use up the free shared allowance this server's finder searches depend on.

Moderation engine

Administration → Moderation brings everything together: the reports people file, the flags raised automatically, the rules that raise them, every hidden post or comment (restorable), and the log of every action with who took it.

RuleWhat it does
Keyword listFlags, hides or blocks posts and comments containing listed words or phrases (* wildcard)
Link limitFlags posts with more than N links
Posting velocityFlags accounts posting more than N times in M minutes
New-account reviewFlags the first posts of accounts younger than N days
Media reviewFlags posts with video (or images) for a human look
  • Decisions: dismiss, hide (reversible), warn the author, suspend for 1 / 7 / 30 days or until lifted.
  • Hiding keeps the content in the database with the reason and the moderator; restoring it is one click under Hidden content.
  • Every decision is written to the audit log and to the moderation action log.

Hikma Research manual · generated from the same content as the online documentation · 2026-11-16