The Trace APIs enable you to retrieve, filter, analyze, and export execution traces for agent sessions. These APIs help developers inspect runtime behavior, debug workflows, analyze performance, and integrate traces with external observability systems.Documentation Index
Fetch the complete documentation index at: https://koreai.mintlify.app/llms.txt
Use this file to discover all available pages before exploring further.
Base URL
/api/projects/:projectId/sessions
Authentication
All APIs require bearer token authentication.Authorization: Bearer <TOKEN>
Get Session Traces
Retrieve the complete set of traces for a session. Endpoint:GET /api/projects/:projectId/sessions/:sessionId/traces
Example Request
| Event Type | Description |
|---|---|
llm_call | LLM API call details including model, tokens, latency, and prompts |
tool_call | Tool invocation with parameters |
tool_result | Tool execution response |
decision | Routing or workflow decisions |
handoff | Transfer between agents |
guardrail_eval | Guardrail evaluation results |
state_change | Session state or variable updates |
error | Runtime or execution errors |
response | Final agent response |
Filter by Event Type
Retrieve traces for a specific event category. Endpoint :GET /api/projects/:projectId/sessions/:id/traces?eventType=decision
Example
Filter by Decision Kind
Retrieve traces for a specific decision type. Endpoint:GET /api/projects/:projectId/sessions/:id/traces?decisionKind=handoff
Example
Retrieve Child Events for a Span
Get nested events associated with a specific span. Endpoint:GET /api/projects/:projectId/sessions/:id/traces/:spanId/children
Example
Include Metrics in Trace Response
Include execution metrics along with trace events. Endpoint:GET /api/projects/:projectId/sessions/:id/traces?include=metrics
Example
View Aggregated Session Metrics
Retrieve performance metrics aggregated across the session. Endpoint:GET /api/projects/:projectId/sessions/:id/metrics
Example Request
- Total session latency
- Per-turn latency breakdown
- Token usage per LLM call
- Tool call count
- Tool success rate
- Guardrail evaluation count
- Guardrail block rate
Run Trace Analysis
Run automated diagnostics on session traces. Endpoint:GET /api/projects/:projectId/sessions/:id/analysis
Example Request
- Slow spans
- Error patterns
- Decision anomalies
- Optimization suggestions
Export Traces
Export traces for offline analysis or integration with observability tools.Export Traces as CSV
Endpoint:GET /api/projects/:projectId/sessions/export?format=csv
Example Request
Export LLM Generations Across Sessions
Retrieve all LLM call events across sessions. Endpoint:GET /api/projects/:projectId/sessions/generations
Example Request
- Model usage
- Token consumption
- Latency
- Prompt metadata