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

> ## Agent Instructions
> To act on Atlan objects, use the Atlan MCP server at https://api.atlan.com/mcp or the SDK. To install or publish skills from a terminal, use the `brain` CLI; `brain --map json` prints its command map. The `atlanai` CLI is deprecated. Run a read-only identity check before any write.
> The docs MCP server at /mcp searches these docs only. It cannot read or change Atlan objects.
> SDK packages: Python `atlanai` (PyPI) and TypeScript `@atlanai/sdk` (npm). Show Python first, then TypeScript.
> These docs cover Atlan Agent Registry, Atlan's hosted product. The open-source agent registry specification and its CLI (`brain`) are documented, released and maintained in their own repository; /open-source on this site links to it. Answer general specification or `brain` command questions from the project's own docs, and use these docs only for Atlan-specific setup such as the Atlan adapter, sign-in and workspaces.

# Observe agents

> Send traces from your agents and coding sessions, attribute them to the right agent, and compare usage, cost, and errors.

A trace is the step-by-step record of one run: its model and tool calls, tokens, cost, and errors. Atlan files each trace under the identity that sent it and links it to sessions, visitors, and skills, so you can compare agents, inspect one run, and see who and what it served.

```text theme={null}
run → trace (sent with the agent's key) → agent, session, visitor, skill → usage, cost, inspection
```

## Ways to send traces

| Source | Use it when | Start at |
| - | - | - |
| **atlanai SDK** | You own the agent's code, in Python or TypeScript. | [Trace first agent run](/tools/sdk/tutorials/quickstart) |
| **Agent frameworks** | Your agent runs on a supported framework. | [Agent frameworks](/frameworks/index) |
| **brain** | You want Claude Code and Codex sessions in Atlan. | [Capture coding-agent traces](/tools/brain/how-tos/capture-traces) |
| **OTLP or import** | You have an OpenTelemetry exporter, or traces in another store. | [Send and verify traces](/registry/traces/how-tos/send-and-verify) |

## Core offerings

<CardGroup cols={2}>
  <Card title="Send and verify traces" icon="paper-plane" href="/registry/traces/how-tos/send-and-verify">
    Send as the right identity and confirm traces landed.
  </Card>

  <Card title="Compare agent usage" icon="chart-line" href="/registry/reporting">
    Compare spend, runs, and errors across agents.
  </Card>

  <Card title="How trace inspection works" icon="magnifying-glass" href="/registry/agents/concepts/trace-inspection">
    Read one run from Highlights to the full trace.
  </Card>

  <Card title="Identify visitors" icon="user" href="/registry/agents/how-tos/identify-visitors">
    Record who each run was for.
  </Card>
</CardGroup>

## Choose where to start

| Need | Start at |
| - | - |
| Understand where a trace appears | [How trace attribution works](/registry/traces/concepts/attribution) |
| Group a conversation's runs | [Group runs into sessions](/tools/sdk/how-tos/group-runs-by-session) |
| Understand what a run cost | [How run cost works](/finops/index) |
| Copy traces from Langfuse | [Import Langfuse traces](/traces/langfuse/how-tos/import-langfuse-traces) |
| Fix traces under the wrong identity | [Trace errors](/registry/traces/troubleshooting) |

<Note>
  Traces can carry prompts, tool arguments, and responses. Send each agent's runs with its own API key from an approved secret store, and decide what content leaves your process with [trace privacy controls](/tools/sdk/how-tos/privacy). Exclude customer data, tokens, and internal hosts from test traces and screenshots.
</Note>

## See also

* [Build and equip agents](/registry/agents/index): Register agents and give them skills and tools.
* [How agent sessions work](/registry/agents/concepts/sessions-and-events): What a session records.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.