> ## Documentation Index
> Fetch the complete documentation index at: https://docs.groundforge.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Traces

> Follow one request from input to final delivery.

A Trace is the timeline for one Agent request.

Open **Traces**, select a request, and expand the meaningful steps.

```text theme={null}
Channel input
├─ Model
├─ Gateway decision
├─ Tool or MCP
│  ├─ input
│  ├─ decision
│  └─ output or not executed
├─ Model
└─ Channel output
```

## Questions a Trace can answer

* What did the user ask?
* Which model and capability steps occurred?
* Did the Gateway Allow, Observe, Review, or Block the call?
* Did the capability actually run?
* What input and output were recorded?
* Was the Agent successful but final Channel delivery unsuccessful?

Select a node to inspect its duration, outcome, bounded input, bounded output, and related identifiers.

Known credential fields are redacted. Do not place secrets in prompts, tool inputs, or custom metadata.

> **Screenshot placeholder**\
> Add `/images/observability/trace-detail.png` showing the Trace list, semantic tree, and selected span detail.

For a guided example, see [Request Tracing](/guides/request-tracing).


## Related topics

- [Traces](/concepts/traces-analytics.md)
- [Discovery and Traces](/cli/observability.md)
- [List Traces](/api-reference/endpoints/list-traces.md)
- [Search Traces](/api-reference/endpoints/search-traces.md)