Agents
Agent activity feed
Recent agent runs with the tools each used and the decision it reached.
Authorization
HTTPBearer AuthorizationBearer <token>
JWT access token
In: header
Query Parameters
limit?integer
Range
1 <= value <= 100Default
25offset?integer
Range
0 <= valueDefault
0agent_name?string|null
channel?|
Match
^(web|voice)$merchant_id?|
Optional merchant to act on. Must be one you are a member of.
Format
uuidHeader Parameters
X-Merchant-ID?string|null
Response Body
application/json
application/json
curl -X GET "https://example.com/api/v1/agents/runs"{ "data": { "avg_duration_ms": 0, "completed_runs": 0, "failed_runs": 0, "items": [ { "agent_label": "string", "agent_name": "string", "channel": "web", "confidence": 0, "created_at": "2019-08-24T14:15:22Z", "decision": "string", "duration_ms": 0, "run_id": "dded282c-8ebd-44cf-8ba5-9a234973d1ec", "status": "string", "title": "string", "tools_used": [ "string" ], "used_fallback": false, "workflow": "string" } ], "top_tools": [ {} ], "total_runs": 0 }, "meta": {}, "success": true}Approve a recommendation and execute it POST
The human approval gate. On approval the Action Agent builds the segment and campaign, the Experiment Agent creates the control/test split, and the campaign is launched with simulated delivery. Requires both campaign:execute and experiment:approve.
Agent trace GET
Human-readable trace: the actions taken, tools called, evidence gathered and decisions reached. Raw model reasoning is never recorded or exposed.