List Agents
List the agents this platform exposes — the routing truth.
Each entry carries capabilities + session.kind (structured routing signal)
plus best_for / not_for / example_tasks hints and the run endpoint.
The run BODY schema lives in the OpenAPI spec, not here. Filter with
?capability= (e.g. code → only agents that can run code).
Authorizations
Query Parameters
Filter to agents that expose this capability, e.g. ?capability=code. The closed set of agent capabilities — the registry's routing TRUTH.
capabilities on an agent is platform-enforced: sending workspace_id to an
agent that lacks files → 400 naming the right agent. Both agents do
browser + files; the only structural difference is code / terminal
(browsercode).
browser, code, terminal, files Response
Successful Response
The registry: every agent this platform exposes.
Not paginated — the agent list is tiny and fully enumerable. A plain array so an LLM can read the whole routing table in one call.