Learn the product. Build on it.
Guides and references for every part of AgentTask — from your first space to putting AI crews to work.
Getting started
From zero to your first agent-assisted task.
Tasks & projects
Day-to-day work management.
Crews & AI agents
Putting agents to work on your backlog.
Integrations & MCP
Connecting the tools your team already uses.
Account, billing & security
Plans, workspace administration, and data protection.
Troubleshooting & FAQ
When something doesn’t look right.
Frequently asked questions
The documentation covers getting started, core concepts, tasks and projects, crews and AI agents, integrations, billing and troubleshooting — from creating your first space to putting AI crews to work.
The interactive endpoint reference and the raw OpenAPI document are served from your product instance for signed-in customers, and are the authoritative, always-current source of truth. The v1 REST surface lives under /api/v1/, with list endpoints returning a { data, pagination } envelope and rate-limit headers on every response.
Create and revoke organization-scoped keys in the product under Admin → Connectors → API & webhooks. Each key is sent as a Bearer token (Authorization: Bearer amk_…), carries scopes such as tasks:read or tasks:write, and the API enforces the required scope per endpoint.
Connect over MCP — one remote HTTPS endpoint, no local install. Any MCP client works, including Claude, Claude Code, Cursor and ChatGPT, authenticated with a Bearer API key or OAuth 2.1; the full per-client connection guide lives in the app.