# Read an LLM trace

Understand provider, token, latency, cost, policy, and refusal details for a model call.

- Category: LLM Proxy
- Last updated: 2026-06-19
- URL: https://help.irongorilla.ai/articles/read-an-llm-trace

## A trace explains a model call

A trace keeps the operational record for one model interaction. Use it to answer who made the request, which provider handled it, what it cost, whether policy allowed it, and what [audit evidence](/articles/export-evidence-as-pdf-markdown-or-json) is available for review.

## Trace fields to inspect

Start with the agent and team that initiated the request, then review the provider, model, latency, token count, and cost. The route or fallback selected by LLM Proxy explains why the request took that path.

The most important fields for governance are the policy decision, any blocked or redacted content, and the correlation IDs that connect the trace to [audit reporting](/articles/build-an-ad-hoc-report) or support evidence.

## Related articles

- [Configure model routing in LLM Proxy](https://help.irongorilla.ai/articles/configure-model-routing-in-llm-proxy)
- [Investigate blocked actions](https://help.irongorilla.ai/articles/investigate-blocked-actions)
- [Export evidence as PDF, Markdown, or JSON](https://help.irongorilla.ai/articles/export-evidence-as-pdf-markdown-or-json)

## Support

Need help? Tell us where you got stuck: https://help.irongorilla.ai/contact