Project tokens and API
Project tokens let external tools work inside one project with scoped rights, including an optional, tightly scoped AI write allowance.
What project tokens are for
Project tokens are for scripts, internal tools, automations, and integrations that need access to one project without using a full user session. Create them from API Tokens in the project menu.
Token rights
- Viewer tokens can read
- Editor tokens can read and write where writes are supported
- Admin tokens can also delete where delete is supported
- Optionally, AI writes: a list of workspaces (records, board, whiteboard, ai) and a list of allowed actions such as scheduling an AI task; empty means no AI actions are allowed
What the API covers
- Rows, custom fields, and project columns, including a project’s status values and progress checklists
- Pings by id, with time, timezone, length, name, and linked AI commands
- Notes, files, and record sharing
- Board messages and Whiteboard boards, items, and comments
- People and companies
- Public intake form submissions
- Realtime project events
API or MCP?
The REST API is for apps, scripts, and deterministic integrations that speak HTTP. MCP is for AI agents such as Claude or Codex that should reason over your records with semantic tools and reviewed writes. Both use the same records and permissions; see MCP and agent connections.
Use the API docs for technical details
The wiki explains the product model. The API docs explain endpoints, request bodies, auth headers, idempotency, row field writes, public record tokens, Whiteboard routes, and current backend behavior.
Safe automation habits
- Create a token only for the project the automation needs
- Use the lowest rights that work, and leave AI writes off unless the integration needs them
- Fetch columns before writing row fields
- Write field values through custom fields, not row metadata
- Rotate or revoke tokens that are no longer needed