The desktop app, screen by screen

Dashboard

Greeting, today's numbers, a 14-day run sparkline, the live activity feed, and the right-hand column: Today (running experiments, checkable due todos, continue-where-you-left-off), Momentum (logbook streak + activity heatmap), your Compass goals, and an optional AI digest.

The ResearchOS dashboard

Overview

The cross-workspace view: per-workspace counts and sync status, a "needs attention" radar (overdue todos, failed runs, stale figures, unsynced changes), global search, and a merged activity feed.

Runs & Analysis

The run board groups running experiments, flow chains and everything else; tiles show status, metrics and age. A run's detail page has live logs, metric charts (including streamed history), artifacts with inline previews and a Use in paper button that maps an artifact to a stable figure name, a provenance graph, linked elements, re-run with edited parameters, diffs against any other run, and an interpretation field for recording what the result means, the decision taken and the next step. Analysis adds leaderboards with a baseline anchor, parameter sweeps, scatter/time charts, and one-click export of any chart as a provenance-tracked figure.

Comparing runs and metric charts in the Analysis view

Runs started from the app use your project's environment: a workspace .venv, conda environment.yml, uv or poetry lockfile is detected automatically. The start-run dialog also has a Jupyter notebook toggle — the notebook executes headlessly with the same capture, and the executed copy, outputs included, becomes an artifact of the run.

On a synced workspace, the start-run and re-run dialogs offer Run on another computer: the request is queued (encrypted, like everything else), a machine of yours that has opted in executes it, and the result appears on the board as an ordinary run — logs and artifact files included. Pending and completed requests show in a strip at the top of the Runs screen. See remote runs for setup and details.

Logbook & Notes

The logbook is a color-coded timeline of typed entries merged with automatic events (runs finishing, artifacts appearing). Templates per kind, day headers, weekly stats, project filters — and sealing for tamper-evident records, with optional RFC 3161 timestamps and a one-click evidence-package export. Checklist lines inside an entry become live action items you can tick in place or file as todos. The timeline button replays the whole project chronologically and drafts a progress report for any date range — ready for your next progress meeting. Sync conflicts never cost you a paragraph: each entry's History button lists the versions kept from sync merges, with one-click restore. Notes are wiki-linked ([[Title]]), back-linked, templated, carry workflow statuses (draft / in progress / to review / final), and connect to runs, references and todos.

The logbook timeline with colored entries

Todos & Agenda

Planning buckets (Overdue / Today / This week / Later), a kanban board view, subtasks, recurring items, natural-language quick add, drag reordering inside lists, and "schedule on agenda". The Agenda merges every workspace, color-coded, with a month view and a week grid where events drag and resize in place, recurring events, reminders, deadline flags from due todos, and ICS import/export.

Graph & Project

The graph shows your workspace as connected knowledge — runs, notes, references, todos, with wiki and provenance edges; filter, search, or focus two hops around any node. Project is the compass: goals, hypotheses with attached evidence (runs and notes, supporting or contradicting), and a decision journal.

The knowledge graph linking runs, notes and references

Library & Desk

BibTeX-native references: add by DOI/arXiv id, import BibTeX, PDFs or a Zotero library; shelves (to read / reading / done), ratings and APA-formatted bibliography export. The Desk reads PDFs in-app with multi-document tabs, highlights that sync as annotations, quote-to-note with page references, and a persistent reading note per paper.

The reference library with PDFs and reading statuses

Paper

Markdown with live preview, outline navigation, section scaffolds (IMRaD, thesis chapter, report), snapshots with diffs, citation autocomplete from your library with a cited/uncited panel, figure insertion from run artifacts with hash-based staleness checking, and PDF/DOCX export.

Two buttons in the header connect the paper to your evidence. Paper assets manages stable figure names: map an artifact from any run's Artifacts tab (Use in paper), then export — every mapped file lands in paper-assets/ under its stable name, with provenance manifests, so your LaTeX or Word document updates in place when you re-run and re-export. Slides generates a Marp or Beamer deck skeleton from recent or pinned runs, one slide per run with a closing provenance table.

Lab (or Duo)

The shared-space tab is named after what you have: Duo for a Pro user's own shared space (up to 3 people), Lab when your account belongs to a lab — a full organization with teams, a member directory and governance, set up with us as part of the Lab plan's onboarding.

In any shared space: member roster with roles, join requests, E2EE chat and share requests, announcements, and one-click fulfill — publish an element from any of your private workspaces into the space. Owners and admins get a team overview of recent activity, and any member can post their progress report into the space with Publish report.

Lab members also get, on the same screen:

  • Directory — the lab's teams, people and team spaces, with a search box over what colleagues chose to publish (interests, hand-picked references). A result is one click from a message.
  • Conversations — end-to-end encrypted direct messages and group chats. Messaging someone from the directory creates the thread; the other side's app accepts automatically — no accept step. Flag a message as a request and it lands in the recipient's cross-lab requests inbox, with a mark-answered action.
  • My lab profile — your opt-in expertise profile: research interests plus references you pick by hand. Empty by default; nothing is ever published from your library automatically.
  • Lab overview & audit (admins only) — sync activity per team and space plus an org-wide audit trail, built from metadata only: content stays end-to-end encrypted, and results reach the overview only through reports members explicitly publish.

Labs on a lab plan can also point every member's AI assistant at a model on the lab's own hardware — see the assistant section below.

AI assistant

The ✨ icon opens a chat grounded in your workspace: multi-turn conversations with history, copyable answers, save-as-note, and "ask about this element" from runs and logbook entries. Any element can be individually locked away from it, and a global default can make everything opt-in.

Three ways to power it, chosen from the key icon in the panel:

  • Your own API key — Anthropic (Claude), OpenAI (GPT), Google (Gemini) or Mistral, one key per provider. The key stays on your machine; prompts go directly to the provider.
  • Plan credits — no key needed. Your question is relayed through your sync server to the model provider, answered, and forgotten — never stored or logged. This is a deliberate, explicit choice; the panel shows your remaining credits (10 questions per month during the beta).
  • Lab server (local model) — if your lab has pointed its plan at a model on lab hardware, this provider appears automatically and becomes the default: prompts never leave the lab network.

Settings

Daemon connection, account & devices (create/login, recovery, passphrase change, device revocation, deletion), plan & usage, interface language (system / English / French), notifications, behavior defaults (start screen, agenda scope, week start), AI visibility defaults with per-element locks, sync per workspace, backups, trash, data management (sizes, export-all, transmission packages, unregister), and the tour replays. The shared-space card here is where you create a Duo space from a workspace or accept invitations; it also states that a full lab — teams, directory, governance — is set up with us as part of the Lab plan (contact@tramea.org).

Three sync controls live here too:

  • Join a synced workspace — "Synced workspaces on this server" lists every space on your account; Join here recreates one on this computer with its full history (the CLI equivalent is ros sync join).
  • Sync project files — per workspace: scripts, data and drafts travel between your computers as encrypted blobs. A computer that joins a workspace where this is on gets it automatically.
  • Allow remote runs on this computer — per machine: opt this computer in as an executor for runs queued from your other devices.
Press Ctrl+K anywhere for the command palette: jump to screens, create anything, search everything.