Documentation
Build with Tobie
Everything you need to connect an agent, call the REST API, and ship an integration your HR and security teams will sign off on.
The guides, the API reference and the Connect steps need a Tobie account: anything generated from the API registry, or that connects an app, needs a name. Sign in or request access.
Get started
QuickstartMake your first call to the Tobie API in five minutes.Connect your AIStep-by-step setup for ChatGPT, Claude, Claude Code, Cursor and VS Code.AuthenticationAPI keys for machines, OAuth 2.1 for people and AI apps. One header.Scopes and permissionsEvery operation names exactly one scope. Here is the full matrix.
Core concepts
Tenants and data classesHow Tobie decides which company, which fields and which people a call can see.ErrorsRFC 9457 problem details, stable codes, and what to do about each.Pagination and searchCursor pagination, filters, and searching without putting data in a URL.Idempotency and concurrencySafe retries with Idempotency-Key, and lost-update protection with ETags.Rate limitsBudgets per principal, the headers that report them, and how to back off.ApprovalsSome writes need a human. How proposals, tiers and approvals work.
Build
MCP serverStreamable HTTP at /mcp: discovery, tools, results and connecting clients.Events and webhooksA feed and a stream of what changed, and signed webhooks.Ask TobieRolling outGrounded benefits answers with citations, in an employee’s own AI app.Partners and leadsInvite onlySend leads and follow them, fenced to your own rows.VersioningWhat changes inside /v1, what never does, and how deprecation works.
Docs for machines
Point your agent at these once it holds a Tobie credential. Generated from the same registry the API runs. llms.txt is public; the rest need a sign-in or an API key.