Get a single agent by ID
Returns the agent record. Owners see the full record (including
published_config). Non-owners with binding receive the record
with published_config stripped. Restricted agents seen by
callers without binding return only public-facing fields.
Authorizations
User-bound credential carrying an identity: either a session JWT
or a user access token (at:*) generated from the user settings UI.
Send as Authorization: Bearer <token>.
Org API keys (iak_*) are not accepted here - they carry
no user identity. Use the x-prismeai-api-key header instead
(see OrgApiKeyAuth).
Path Parameters
64Response
Agent record (Agent or AgentRestrictedView depending on caller's access).
- Option 1
- Option 2
Agent record (full view). Owners see this complete record including
published_config. Non-owners with binding receive the same shape with
published_config stripped server-side. Restricted-agent rows
additionally carry access_status (granted or none) when seen in
list/discovery views.
Human-in-the-Loop configuration. Sets a default policy plus optional per-tool overrides. Policies:
auto— run immediately (default).always_ask— halt every call on approval.ask_external— halt only on MCP tool calls.ask_first— halt the first time a tool runs in a session; cache keyed per child name unless an MCP parent rule is configured (then cached per parent).
Controls whether the model can use a given tool on a given turn.
user_first tools are replaced by a same-named, schema-less stub with
a "disabled" description on each turn, unless the user explicitly
summons the tool from the chat input via metadata.tool_choice.
Eligible tool types: function and mcp. file_search, skill,
guardrail, and system types are exempt and always behave as auto.
public, private, restricted draft, published simple, workflow, agent_light, agent_full, orchestrator Snapshot of runtime fields at last publish (owners only).
Only present on restricted-agent rows in list/discovery views.
granted, none