Agents · Update an agent’s configuration
Agents · Update an agent’s configuration
Partial update of an agent’s own fields — title, model, temperature, prompt,
extra.
extra is replaced wholesale rather than merged, so a partial extra
silently drops the keys it omits. Read the current value before writing.
Attachments are not editable here; skills, toolkits and triggers each have
their own endpoints. Toolkits are stored inside extra, which makes that
read-modify-write a trap: a toolkits key sent here is dropped, and the
response still carries the stored value, so the call reads as success.
resource: agent://build-guide covers the toolkit endpoints.
Authentication
Header authentication of the form Api-Key <token>
Bearer authentication of the form Bearer <token>, where token is your auth token.
Path parameters
Agent UUID or slug — both are accepted. From agents_listAgents.
Request
Provider-qualified model, e.g. openai:gpt-4o-mini. Omit to take the platform default rather than guessing a string.
The agent’s standing instructions. Durable behaviour belongs in skills instead — they are editable independently and do not require rewriting the prompt.
draft until you are ready. Only active agents can be run, so a trigger on a draft agent fires and fails.
Built-in tool slugs. Write-only convenience — replaces the whole set, it does not append.
Agent-level flags. allow_skill_override is the one that matters: without it, built-in skills cannot be edited. Toolkits also live in here, but a toolkits key sent with a write is ignored — they have their own endpoints.
