Tracing Servers
Connect observability and monitoring for agent execution
Tracing Servers provide observability into agent execution and performance.
What are Tracing Servers?
Tracing servers collect telemetry data from agent execution:
- Traces - End-to-end request flows
- Spans - Individual operation timing
- Logs - Structured event data
- Metrics - Performance statistics
Why Tracing?
Understanding agent behavior is critical for:
- Debugging - Find why an agent failed
- Performance - Identify slow operations
- Cost - Track token usage
- Compliance - Audit agent decisions
Supported Platforms
Langfuse
Open-source LLM observability platform:
- Trace visualization
- Cost tracking
- Prompt management
- Evaluation metrics
OpenTelemetry (Soon)
Standard observability protocol:
- Vendor-agnostic
- Wide ecosystem support
What Gets Traced
| Event | Data Captured |
|---|---|
| Agent Turn | Input, output, duration, tokens |
| Tool Call | Tool name, args, result, timing |
| Memory Search | Query, results, relevance |
| Context Build | Template, data, token count |
Configuration
Connect your tracing server in team settings:
- Add tracing server connection
- Provide API key and endpoint
- Select which teams to trace
- View traces in your observability platform