Science Without Walls
Local-first, model-agnostic AI research workbench. Run the entire research loop — from exploration to publication — in one auditable desktop session.
7
Languages
3
Platforms
v0.3.4
Latest
Local-first, model-agnostic AI research workbench. Run the entire research loop — from exploration to publication — in one auditable desktop session.
One meta-skill drives four stages — each produces real, inspectable artifacts, not just chat replies.
Turn a broad research direction into concrete topics, hypotheses, and a pre-survey.
Produce a 6–20 page PDF literature review with 60+ real, verified citations.
Design docs, runnable code, results with provenance, figures, and analysis reports.
8–14 page research paper, 200+ citations, 4–8 figures — publication-grade output.
Every feature designed to accelerate scientific work while keeping you in full control.
The bundled ai4s-agent chains specialist skills end to end — explore, survey, experiment, write — producing inspectable artifacts at every stage.
Every figure, table, and report links back to the exact code, inputs, environment, model output, and conversation that produced it.
Sessions, data, provenance, notebooks, and run records live in local folders on your machine. Nothing leaves by default.
Bring your own model. The runtime talks through a clean SDK layer — providers, skills, and MCP servers stay pluggable.
Local, SSH/Slurm, Modal, and notebook-batch runs are captured as reproducible run records with full environment snapshots.
A token-authenticated gateway serves the real UI to your phone or LAN browser. Start a run at your desk, check results on the go.
Drive your own Chrome — profile and logins intact — to read the live web the way you do, or an isolated private browser on demand.
Agent skills, MCP servers, one-click science connectors (arXiv, PubMed, Crossref, Materials Project…), slash commands, and shell mode.
A comprehensive platform that covers the entire scientific workflow.
Run multiple research sessions simultaneously with isolated workspaces, shared file system, and full history.
Real .ipynb files with Python and R kernel execution, managed Jupyter environment, and seamless agent integration.
PDF, images, video, HTML, Markdown, CSV/TSV tables with charts, molecules, 3D meshes, genome tracks, and more.
One-click access to arXiv, PubMed, Crossref, Semantic Scholar, bioRxiv/medRxiv, and more research databases.
Built-in auditor checks images, numbers, logic, and citations with 4-level evidence grading before publication.
Command execution, file deletion, dependency install, and remote connections all require explicit human approval.
Decoupled layers with a strict SDK boundary — the UI never calls the agent runtime directly.
Free, open-source, available for all major platforms.
Apple Silicon & Intel, macOS 13+
Coming Soonx86_64 .deb / .rpm
Coming Soon