Local-first, model-agnostic AI research workbench for macOS, Windows & Linux.
Formerly Open Science. An open-source desktop alternative to Claude Science and similar AI-for-science workbenches — built with Tauri, MCP, agent skills, and reproducible artifacts. It connects agents, notebooks, files, figures, reports, runs, and review into one auditable desktop workflow.
English · 简体中文 · 日本語 · Español · Deutsch · Français · 한국어
- 2026-08-01 — 🗂️ Projects, memory, and full history. Group sessions into named projects (import an existing repo in place, no copying), give the agent persistent global and project memory, and reach every past conversation through a searchable history with archive, restore, and export. (v0.3.1)
- 2026-07-24 — 🪟 Split-pane tiling. Tile sessions side by side, drag panes to re-dock them, keep several independent Screens, and run a different model in each pane. (v0.3.0)
- 2026-07-21 — 🌐 Access from anywhere — even your phone. A token-authenticated gateway serves the real desktop UI to a CLI, a browser on your LAN, or your phone (loopback by default; LAN is opt-in). Start a run at your desk and read the finished figure and report on your phone. (v0.2.3)
- 2026-07-21 — 🧭 Browser control. The agent can drive your own Chrome — profile and logins intact — to read the live web the way you do, or an isolated private browser on demand. (v0.2.3)
- 2026-07-09 — 🎉 #1 on ResearchClawBench. Open Science Desktop ranks #1 by scored-task average on ResearchClawBench, an end-to-end benchmark for autonomous scientific research agents (Pass@1 leaderboard).
- ✨ What it does
- 🎬 See it in action
- 🧪 Current capabilities
- 🔌 Skills and connectors
- 📦 Install
- 🚀 Build from source
- 🔒 Safety and privacy
- 🗂️ Repository layout
- 📌 Status
- 🤝 Contributing
- 📖 Citation
- ⚖️ License
Runs the whole research loop — from a broad direction to a finished paper: exploration, literature survey, hypothesis, experiment code, analysis, figures, and write-up, in one continuous, auditable session.
- Autonomous research agents — the bundled
ai4s-agentchains specialist skills end to end (explore → survey → experiment → write), and each stage drops a real, inspectable artifact into your workspace, not just a chat reply. - Everything traces back — figures, tables, reports, notebooks, and run outputs link to the exact code, inputs, environment, model output, and conversation that produced them.
- Local-first and yours — sessions, data, provenance, notebooks, and run records live in local folders on your machine. Nothing leaves by default.
- Model-agnostic runtime — the UI talks through
packages/sdkto a bundled, pinned OpenCode sidecar. Bring your own model; providers, skills, and MCP servers stay pluggable. - Reproducible by construction — local, SSH/Slurm, Modal, and notebook-batch runs are captured as reproducible run records, not loose terminal scrollback.
- Reach it from anywhere — a built-in, token-authenticated gateway serves the real desktop UI to a browser on your LAN or phone (or, with a tunnel, from anywhere) — kick off a run at your desk and check on it from your phone over lunch. Off by default; loopback-only until you opt in, and API keys never leave the machine.
- Drives your own browser — the agent can control your real Chrome, with your profile and logins intact, to read the live web the way you do — or an isolated private browser when you'd rather it not.
- Plan before it acts —
/planlays out an execution plan before touching a file, and/goalfixes the objective, constraints, and acceptance criteria the agent then works toward. - Built for long projects — named projects group their sessions, two layers of persistent memory (global and per-project) carry what matters between them, and a long conversation compacts itself as it approaches the model's context window.
- Work several threads at once — tile panes side by side, keep independent Screens, and give each pane its own model.
- Extensible — agent skills, MCP servers and one-click science connectors,
/commands,!shell mode, and a model-agnostic SDK.
One prompt -> a publication-grade figure, and every point traces to the exact code and inputs that made it. No black boxes: open any artifact to see its generating script, its data files, and the conversation that produced it.
Literature -> a verifiable report. Fan the search out across sources, draft a manuscript rendered as a PDF, and gate it on a citation review — DOIs resolved, unsourced numbers and figure/code inconsistencies flagged — before anything ships.
Drives your own Chrome. The agent reads the live web through your real browser profile — logins and all — then turns what it finds into a figure and a sortable CSV.
Research from anywhere — even your phone. A built-in authenticated gateway serves the real desktop UI to a browser on your LAN (or a tunnel), so you can kick off a run at your desk and read the finished figure and report on your phone.
 New session |
 A finished analysis |
 A reproduced benchmark |
More screenshots
The research loop, as skills. One meta-skill runs the full pipeline; each stage is a self-contained skill that produces a real, gradeable artifact — runnable on any model OpenCode supports:
| Skill | Role | Primary output |
|---|---|---|
ai4s-agent |
Runs the four skills below, in order | The full research package |
research-explorer |
Turn a broad direction into concrete topics | research_exploration.md, topic_matrix.md, literature_pre_survey.md |
literature-survey |
Write a literature survey | 6–20 pp PDF, 60+ real citations, LaTeX source, taxonomy figures |
experiment-suite |
Build an experiment package | Design doc, runnable code, results.json with provenance, figures, report |
paper-writer |
Write a research paper | 8–14 pp PDF, 200+ citations, 4–8 figures, tables |
mindmap-render |
Render a mindmap | Image generated from a topic_matrix.md |
integrity-auditor |
Audit a paper's integrity | Image / numerical / logical findings, 4-level evidence grading, audit_report.md |
These ship in the ai4s-skills pack alongside first-party review skills and the
office/document skills below.
| Area | Current state |
|---|---|
| Desktop shell | Tauri 2 + React + TypeScript + Vite, with macOS, Windows, and Linux desktop builds. |
| Runtime | Bundled OpenCode sidecar, auto-started by the app, isolated from the user's own OpenCode config/data. |
| Projects | Named project workspaces that group their sessions; import an existing folder in place (never copied) or adopt one already inside the workspace; move an existing session into a project. |
| Sessions | Multi-session chat/history, dated workspace folders, searchable history with archive/restore/export, @ file and # conversation references, / commands, and ! shell mode. |
| Layout | N-ary split-pane tiling with drag-to-dock, independent Screens, per-pane model and reasoning effort, and cross-screen pane drag. |
| Agent modes | /plan for plan-then-execute, /goal for objective and acceptance criteria, live subagent status in its own panel, and Stop that reflects the runtime's real server state. |
| Memory | Global and per-project memory layers, switchable, plus automatic context compaction as a conversation approaches the model's window. |
| Remote compute | Register machines from your ~/.ssh/config, probe them, and submit, track, or cancel jobs from the app. |
| Appearance | Light, Warm, and Dark themes with per-theme accents, and UI zoom. |
| Files | Global and per-session file browsing, context menu actions, external open/reveal, copy path, and local preview server. |
| Remote access | Token-authenticated gateway that serves the real UI to a CLI, a LAN web browser, or your phone (loopback by default, LAN opt-in); read-only vs full access modes; copy a link with the token embedded to connect in one tap. API keys never cross the wire. |
| Browser control | The agent drives your own Chrome — profile and login state preserved — reading pages through the accessibility tree, or an isolated/private browser on demand. |
| Notebooks | Real .ipynb files, Python and R notebook creation, local kernel execution, managed Jupyter environment via bundled uv, and an Open JupyterLab action. |
| Runs | Append-only run logs, global SQLite run index, search/facets/pagination, local/remote surfaces, output links, logs, and reproduce prompts. |
| Provenance | .openscience/provenance.jsonl tracks file versions and links produced artifacts back to the run or edit that created them. |
| Review | Traceability, statistics-integrity, domain-check, large-file, publication-figure, remote-compute, and Modal run skills are bundled as first-party skills. |
| Viewers | PDF, image, video, HTML, Markdown, code, CSV/TSV tables with charts, DOCX, XLSX, PPTX, molecules, 3D meshes, genome tracks, FITS, DOS/DOSCAR, EIGENVAL bands, qcode, anomaly maps, and phase files. |
| Models | OpenCode provider catalog, OAuth/API-key provider flows, custom OpenAI-compatible endpoints, and local/provider-specific options supported by OpenCode. |
| Interface languages | English, Simplified Chinese, Japanese, Spanish, German, French, and Korean. Portuguese (Brazil) and Arabic are registered but not selectable yet. |
Bundled skills are fetched for builds and releases instead of being committed into git history:
ai4s-skillspack fromai4s-research/ai4s-skills.- Office/document skills from the Apache-2.0
anthropics/skillsrepository:docx,pdf,pptx, andxlsx. - First-party core skills in
runtime/skills/core/:traceability-review,stats-integrity,domain-check,large-file,publication-figures,remote-compute, andmodal-run.
One-click science MCP connectors currently include:
- Literature search: arXiv, PubMed, Crossref, Semantic Scholar, bioRxiv/medRxiv.
- Biomedical databases: PubMed, ClinicalTrials.gov, MyVariant/ClinVar.
- Materials Project.
- FRED economic data.
- Space weather.
- Open-Meteo weather and climate.
- USGS water data.
You can also add any local or remote MCP server from Settings. See
docs/CONNECT_YOUR_TOOLS.md.
For a neutral positioning note, see
Open Science Desktop vs OpenScience.
Download the latest installer from the Releases page.
- macOS:
.dmg/.app, Apple Silicon and Intel, macOS 13 Ventura or later. - Windows: NSIS
.exeand.msi, Windows 10/11 x64. - Linux:
.deband.rpmon x86_64 Linux.
The macOS packages are Developer ID signed, notarized, and stapled, so they open
normally — no xattr workaround needed. Windows and Linux builds are not signed yet.
Windows: if SmartScreen appears, choose More info -> Run anyway.
Linux:
sudo apt install ./Open.Science_*.deb
# or
sudo rpm -i Open.Science-*.rpmPrerequisites:
- Node.js >= 20
- pnpm 9
- Rust toolchain
- macOS, Windows, or Linux system dependencies required by Tauri
git clone https://github.com/ai4s-research/open-science
cd open-science
pnpm install
# Fetch pinned sidecars and bundled skills. These are git-ignored.
bash scripts/dev/fetch-opencode.sh
bash scripts/dev/fetch-uv.sh
bash scripts/dev/fetch-skills.sh
# Run in development or build installers.
pnpm --filter @ai4s/desktop tauri dev
pnpm --filter @ai4s/desktop tauri buildUseful checks:
pnpm test
pnpm typecheck
pnpm lint- Workspace files, raw data, session history, provenance, notebooks, and run records stay local by default.
- Command execution, file deletion, dependency installation, and remote connections are human-approved flows in the desktop app.
- Provider credentials are written to app-private runtime config, not to the workspace, provenance, git, exports, or global OpenCode config.
- Settings includes a plain-language data-flow view explaining what can be sent to the selected model provider.
| Path | Purpose |
|---|---|
apps/desktop/ |
Tauri + React desktop app. |
packages/sdk/ |
OpenCodeClient; keeps the UI from calling OpenCode directly. |
packages/shared/ |
Shared domain types and chart palette. |
packages/ui/ |
Shared UI package. |
runtime/skills/core/ |
First-party scientific skills. |
runtime/skills/external/ |
Build-fetched external skills. |
runtime/harness/ |
Runtime harness knowledge and operator context. |
runtime/mcp/ |
MCP runtime notes/configuration. |
examples/ |
Built-in example workspaces. |
scripts/dev/ |
Sidecar, uv, skill fetchers, and focused regression probes. |
docs/ |
Product, technical, operator, connector, and research notes. |
The project is a working desktop MVP in active development. The most reliable current
implementation log is PROGRESS.md. Product and architecture notes
live in docs/PRD.md and
docs/TECHNICAL_DESIGN.md, but those documents include
target design as well as historical status notes.
Near-term work is focused on Windows code signing, auto-update, broader Windows/Linux verification, richer connector hardening, and continued reproducibility review. macOS releases are already signed and notarized.
Issues and PRs are welcome. Keep changes minimal and verifiable, follow
AGENTS.md, and run the checks before opening a PR. For discussion,
join the Open Science Discord or the
linux.do community.
If you use Open Science Desktop in your research, please cite it:
@software{open_science_desktop,
author = {{The Open Science Desktop Contributors}},
title = {Open Science Desktop: a local-first, model-agnostic AI research workbench},
year = {2026},
version = {0.3.3},
doi = {10.5281/zenodo.21805331},
url = {https://github.com/ai4s-research/open-science},
license = {MIT}
}GitHub's "Cite this repository" button (top of the repo page, generated from
CITATION.cff) provides the same reference in APA and BibTeX.
MIT. Bundled third-party skills and connectors keep their own licenses.
Open Science Desktop is beta research tooling. Treat outputs as drafts: verify numbers, citations, code, and conclusions before publication or decision-making.
