Agent Endpoints
Agent execution and management endpoints. Served through the API gateway at http://localhost:8000.
Endpoints
| Method | Path | Description |
|---|---|---|
POST | /execute | Execute an agent pipeline |
GET | /conversations/{id}/stream | Stream agent response via SSE |
POST /execute
Execute a full agent pipeline synchronously. The agent runs through the complete LangGraph StateGraph: system prompt, memory retrieval, LLM call, tool execution, skill injection, HITL approval, synthesis, and output formatting.
Request:
curl -X POST http://localhost:8000/execute \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{
"agent_type": "rule_37",
"user_message": "I need a spacing exception for well API 42-383-12345 on the Smith Ranch lease",
"conversation_id": "conv-optional-id",
"context": {}
}'| Field | Type | Required | Description |
|---|---|---|---|
agent_type | string | Yes | One of: rule_37, rule_32, compliance_monitor, flaring_monitor |
user_message | string | Yes | The user’s input message |
conversation_id | string | No | Existing conversation ID, or auto-generated |
context | object | No | Additional context to pass to the agent |
Response (200):
{
"conversation_id": "conv-abc123",
"agent_type": "rule_37",
"response": "Based on the spacing analysis for well 42-383-12345...",
"tools_called": ["spacing_calculation", "offset_well_analysis"],
"skills_injected": ["spacing-calculation"],
"approval_required": true,
"token_usage": {
"prompt_tokens": 4200,
"completion_tokens": 1800,
"total_tokens": 6000
},
"cost_usd": 0.12
}GET /conversations/{conversation_id}/stream
Stream agent responses in real-time via Server-Sent Events. See the SSE Events page for the full event type reference.
Request:
curl -N -H "Authorization: Bearer $TOKEN" \
http://localhost:8000/conversations/conv-abc123/streamResponse: text/event-stream with JSON event objects.
data: {"event_type": "agent_status", "status": "thinking", "node": "system_prompt_node"}
data: {"event_type": "token", "content": "Based on "}
data: {"event_type": "token", "content": "the analysis..."}
data: {"event_type": "done", "final_response": "..."}The SSE endpoint is the recommended way to interact with agents in production. The synchronous /execute endpoint is useful for testing and batch operations.
Last updated on