Documentation
What Daslab is, how to set your first scene, and the concepts underneath — every page, grouped the way the sidebar reads.
Start here
- What is Daslab? — A workspace where AI agents do real work in the systems you already run — scenes, cells, approvals, history.
- Get started — Run your first scene in five minutes.
- Your first blueprint — Write a scene as JSON and push it live in five minutes.
- Daslab CLI and MCP — Drive Daslab from Claude, Claude Cowork, Claude Code, Cursor, the shell, or any MCP client.
- Vision — The product is one slice. The north star is a self-driving lab — autonomous research on physical reality, tested in Bangkok.
Concepts
- How Daslab works — What Daslab is and how it works — scenes, cells, agents, approvals, history.
- Scenes — Persistent shared workspaces backed by live conversations.
- Assets — The things you govern — accounts, records, notes, programs. Identity, gates, audit, blueprints.
- Cells — How a scene shows assets — refs plus a projection. Rendering, sizes, views, and what the agent sees.
- Knowledge Bases — Notes with folders and links, in the scene where the work happens — searchable by you and by the agent.
- Agents — The AI layer. Models, tools, and the loop that calls them.
- Jobs — A single conversation that ran — the prompt, the tool calls, the result, and the trace.
- Programs — The agent's plan as an artifact — ops with explicit data flow, saved by name, run again with new inputs.
- Approvals — Every write is a confirmable action. Reads are open; writes are intentional.
- Sandbox — When no tool fits, the agent writes code and runs it in an isolated sandbox. What the Sandbox can do and what it can't touch.
- Providers — Where asset types come from. The integration layer. Every connected service is a provider.
- Blueprints — Scenes authored as data — apply, fork, merge, push, with credentials as sealed references.
- Provider Scenes — Providers authored as data, not code. The kernel that lets a scene act as a provider.
- Commit chain — Every scene change is a commit — content-addressed, tree-hashed, diffable. One chain, per scene.
- Traces — Every job leaves an OpenTelemetry trace — the record of what ran, what it cost, what came back.
- Audit — The reading side of the records — what traces and the commit chain log, who may read them, how long they're kept.
Operating
- Auth & permissions — OAuth for connections, SSO for orgs, permissions inherited from the systems you connect.
- BYOK (bring your own keys) — Managed usage by default — one bill, hard caps. On Enterprise, run on your own provider accounts.
- Data, residency, retention — Where your data lives, how long it stays, and what we never do with it.
Reference
- API console — Interactive OpenAPI reference for the REST API.
- API v1 reference — The current API surface, including OpenAI-compatible jobs.
- Self-hosting — Run your own instance — one binary, a Postgres, a Redis.
Influences
- Influences — The thinking behind Daslab's design — what we read, what we steal, what we explicitly reject. Sources we keep returning to.
Changelog
- Docs site, take one — We rebuilt the website around docs. Here's why and what's next.