# What is agent observability?

> Agent observability is the practice of tracing and monitoring the full agent harness: LLM calls, tool calls, MCP requests, retrieval, multi-step trajectories, cost, latency, and outcomes.

- HTML: https://openlit.io/glossary/agent-observability
- Markdown: https://openlit.io/glossary/agent-observability.md

LLM observability usually focuses on individual model calls. Agent observability covers the trajectory: which tools ran, what context was passed, where the agent looped or stalled, and what the user-facing outcome was.

OpenTelemetry GenAI semantic conventions give a portable way to represent those spans. OpenLIT emits standard OTLP traces and metrics so the same data can power OpenLIT dashboards or export to Grafana, Datadog, and other OTLP backends.

Coding agents and MCP servers are part of the same picture: sessions, tool calls, tokens, and cost per user or repo without requiring a proxy.


## Related

- [Agent harness](https://openlit.io/glossary/agent-harness.md)
- [Agent evals](https://openlit.io/glossary/agent-evals.md)
- [Trajectory evaluation](https://openlit.io/glossary/trajectory-evaluation.md)
- Pillar: https://openlit.io/agent-harness-engineering.md