Browse the documentation
Documentation
Snaga documentation
Learn how to set up an AI company, organize its teams and control their work. For integrations and applications, explore the API and SDK guides.
QuickstartFrom sign-up to a first finished task in the web app — your company, its Chief of Staff, and one piece of work done.REST API referenceEvery endpoint, generated from the OpenAPI 3.1 document the server publishes.Official SDKsTyped clients for TypeScript, Python, Rust, Kotlin, Swift and Ada, generated from the same API description.Chat Completions APIPoint an existing SDK at Snaga without changing the request shape.
Getting started
How Snaga works, your first finished task, and your first call to the API.
- How Snaga worksLearn how a company, its agents and assignments fit together, and which decisions need your approval.
- QuickstartFrom sign-up to a first finished task in the web app — your company, its Chief of Staff, and one piece of work done.
- Your first API callCreate an API key, call the Snaga API with curl, give the Chief of Staff a task and read the result.
Platform concepts
Companies, agents, the Chief of Staff, conversations, runs, memory and files.
- Companies and membersA company is the workspace everything else lives in; members join it by invite and act in it through a role.
- AgentsHow agents are created, changed, versioned, paused and removed, and how the platform chooses the model they run on.
- The Chief of StaffThe head agent every company has — it turns what you ask into work for the right agents, and it can bring in an agent you already use.
- SessionsA session is one conversation with one agent; each message you send starts a run, and files can travel with it.
- RunsA run is one piece of work an agent does — how to start one, follow it live, answer it when it stops for you, and pick it up again.
- Memory and knowledgeWhat an agent remembers between runs, what it always keeps in mind, and the shared documents it can search.
- Files, workspaces and projectsWhere documents live on Snaga — files you hand to an agent, workspaces agents work in, and projects that keep related chats together.
Automation
Work that runs on a schedule, missions with a goal, and webhooks that tell you what happened.
- SchedulesHave an agent start work on its own at set times — a morning digest, a weekly report — with a cron line and a time zone.
- MissionsTurn a larger goal into objectives that agents carry out in order, with budgets, checks, an optional go-ahead from a person and a review at the end.
- WebhooksLet Snaga tell your systems when something happens — a run finished, a budget was crossed — with signed HTTP deliveries you can verify.
Tools and connections
Connect the services you use, MCP in both directions, and agents that talk to other agents.
- IntegrationsConnect the services your company already uses — GitHub, Google Drive, HubSpot, Telegram and more — and choose which agents may use each one.
- MCPGive agents tools from Model Context Protocol servers, and reach your own agents from any MCP client.
- A2ALet other agents discover your published agents and hand them work over the Agent-to-Agent protocol.
Control and safety
Approvals, the rules agents work under, and how usage and limits are counted.
- ApprovalsHow an agent stops and asks before it acts, how you answer — allow, decline with feedback, or refuse — and what happens if nobody does.
- Governance and guardrailsThe limits that hold whether or not anyone is asked — pausing the company, per-run caps, agent permissions, guardrails and the company's rules.
- Usage and limitsPlans and their quotas, how to read where your company stands, your own spend cap, and how to tell a short throttle from a wall that needs a plan change.
Using the API
Authentication, requests and errors, pagination and streaming, and the compatible endpoints.
- AuthenticationAPI keys and their scopes, the Bearer header, safe retries with Idempotency-Key, and how the API is versioned.
- Requests and errorsWhat a request must look like, how a refusal is reported as a problem document, rate limits, and which failures are worth retrying.
- Pagination and streamingWalk long lists with opaque cursors, and follow runs, sessions and tasks live over Server-Sent Events.
- Compatible endpointsConnect clients to Snaga — chat completions and responses run your agents; the Messages API uses your available models.
SDKs
Official client libraries: install, authenticate, stream, page and handle errors.
- SDKs overviewSix official client libraries for the Snaga API, what they share, and how to install each one.
- Client and authenticationCreate a client in each SDK, give it an API key, and set the base URL, timeout and retries.
- Streaming and paginationWalk every page of a list with listAll, and follow a run's live events with a stream that reconnects on its own.
- Errors and retriesHow the SDKs turn problem documents into typed errors, which failures they retry, and how idempotency keys make writes safe to repeat.