Agent API

Call the agent from your own scripts, CI jobs and tools over HTTP, and get the answer when the turn finishes or streamed as it happens.

Everything a run does is an ordinary conversation turn: the same tools, permissions, Projects and usage accounting as a message you type in the web app. Every path goes through the address you open Chetty on, under /api.

The developer path

Read these in order; each one builds on the one before.

1. Create and manage a personal API token: the credential every script, client and tool below uses.
2. Send your first agent API request: the header, one run and its answer.
3. Stream an agent reply, Continue a conversation through the API and Choose a model through the API: the rest of running the agent.
4. Cap what an API token can run and Agent API errors: limits and every refusal.
5. API reference: every other endpoint your account can call on this instance, including plugin endpoints, with their inputs and ready-made examples.
6. Use Chetty as an MCP server for an MCP client, and Run the CLI from scripts for the terminal app; both reach the same endpoints.
7. For plugin authors, Connecting plugins shows how one plugin uses another plugin's connection.

Personal API tokens

Every script, client and tool here signs in with a personal API token that acts as you. Create and manage a personal API token is the one guide for creating, using, limiting and revoking one.

Create a token

Follow Create a token.

Usage and limits

See Review usage and limits, and Cap what an API token can run for the four caps.

Delete a token

See Revoke a token.

What a token cannot do

See What a token cannot do.

Authentication

See Before you start for the header every request carries.

Run the agent

Send your first agent API request runs one turn and reads its answer; the API reference lists every field the request accepts.

Wait for the answer

See When the call waits too long.

Stream the turn

See Stream an agent reply.

Conversations

See Continue a conversation through the API.

Models

See Choose a model through the API.

List the models you can use

See List the models you can use.

Pick a model for one run

See Pick a model for one run.

Switch a conversation's model

See Switch a conversation's model.

Errors

See Agent API errors for every status code and body.

Limit refusals

See Limit refusals.