Agent Telemetry

What is Agent Telemetry?

Agent telemetry provides performance metrics and execution traces for your agents through OpenTelemetry integration. Monitor how your agents perform, where time is spent, and identify optimization opportunities.

Production Sessions and Analytics

The Agentuity platform provides comprehensive telemetry for deployed agents through the Sessions interface. All production agent executions are automatically tracked and persist for analysis and troubleshooting.

Session Overview

The Sessions dashboard displays:

• Status indicators: Success, failure, and execution state
• Agent identification: Which agent handled each request
• Performance metrics: Duration and cost per session
• Timeline data: When sessions occurred with precise timestamps

Detailed Session Analysis

Each session provides detailed execution traces with comprehensive telemetry data:

Understanding Spans

Telemetry data is organized into spans representing units of work. The timeline visualization uses color coding:

• Root (Black): Overall agent execution
• API/HTTP (Teal): External service calls
• AI Services (Purple): Language model interactions
• Vector (Light Blue): Vector database operations
• KV (Orange): Key-value storage operations
• LLM (Pink): LLM-specific operations
• Other (Gray): Miscellaneous operations

Each span includes:

• Start and end timestamps
• Duration in milliseconds
• Relevant attributes (model used, tokens consumed, etc.)
• Parent-child relationships

Viewing Span Details

Click on any span in the timeline to view detailed information in the sidebar:

Performance Analysis

Use telemetry to identify bottlenecks:

1. Long-running spans: Operations taking excessive time
2. Sequential operations: Tasks that could run in parallel
3. Repeated calls: Opportunities for caching
4. Token usage patterns: Optimize prompt efficiency

OpenTelemetry Integration

Agentuity uses OpenTelemetry standards, enabling:

• Export to external observability platforms
• Custom span creation for business logic
• Correlation across distributed systems
• Industry-standard tooling compatibility

For implementation details and code examples, see our SDK documentation for JavaScript and Python.