Run teem from an agent or your own code
Everything the dashboard does is available over REST and MCP with a token you create.
There is nothing in teem that only the dashboard can do. Replying to a conversation, triaging a request, writing a changelog entry, publishing a help article — all of it is available over a REST API and as MCP tools.
Create a token in Settings → API & MCP. It belongs to you and to one organization, and you can see when it was last used.
Which one to use
REST if you are writing code — a script, an internal tool, a job that posts your release notes when you deploy.
MCP if you are pointing an AI agent at teem. The tools are described well enough that an agent can work out which one it needs, so you can say "reply to the newest conversation" rather than reading documentation about it.
Both use the same token and the same permissions as the person who created it. A token made by a member cannot do owner things.
Recipes: hand these jobs to an agent has three jobs written out step by step.
Keeping the token safe
Treat it as a password. It goes in server-side configuration or your agent's local settings, never into a repository, a browser, or a file your project commits. If a token leaks, delete it on that screen and make a new one; nothing else has to change.
This is a different key from the two on the Keys screen. The organization key is public and identifies your app. The signing secret proves who your users are. Neither of them can read your inbox, and this one can.
Where the reference lives
The API reference lists every REST endpoint and who can call it. The MCP reference covers connecting and every tool; hand it to an agent and it can set itself up. When a call fails, My agent gets an error from teem explains the codes.