> ## 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 `aireg` CLI; `aireg --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 Trident, Atlan's hosted product. AI Registry and Agent Registry are its earlier names. Registry OSS is Atlan's open-source agent registry: the registry specification and its aireg CLI, with their own public repository, docs and release notes; /open-source on this site links them. Registry OSS is open source; Atlan Trident is not. Answer generic aireg or specification questions from the project's docs, and use these docs for Atlan-specific setup such as the Atlan adapter, sign-in and workspaces.
> Atlan offers a second product, the Agentic Data Catalog: Atlan's enterprise context layer, covering the data catalog, connectors, lineage, data governance and glossary. It is documented at https://docs.atlan.com; answer questions about it from those docs, not these.

# Monitor agents

> See what your agents did, test whether a change made them better, and find what their runs cost.

These pages are for teams that run agents and need to see what each run did, how well it did it, and what it cost. Use them to find a failing step, prove a change helped, and explain a spend increase.

```mermaid theme={null}
flowchart LR
  A["Trace runs"] --> B["Inspect a run"]
  B --> C["Add the case to a dataset"]
  C --> D["Run an eval"]
  D --> E["Release the change"]
  E --> F["Score production sessions"]
  F --> A
```

A [trace](https://platform.atlan.com/registry/traces/index) is the step-by-step record of one run: model calls, tool calls, tokens, cost, and errors. A [session](https://platform.atlan.com/concepts/how-sessions-and-traces-work.md) ties a run, or a conversation's runs, to the agent that ran it. An [eval](https://platform.atlan.com/concepts/how-evals-work.md) runs a set of cases and gives each case a score, and a [scorer](https://platform.atlan.com/evals/concepts.md#scorer), your code or a hosted judge, decides what the score means. [Cost](https://platform.atlan.com/finops/concepts/how-cost-is-calculated.md) is the recorded tokens of each model call, priced at your organization's rates.

## Core offerings

> [card] [Record agent runs](https://platform.atlan.com/registry/traces/index)
>
> Send traces and group runs into sessions.

> [card] [Inspect one run](https://platform.atlan.com/registry/traces/how-tos/inspect-runs.md)
>
> Follow model calls, tool calls, and errors step by step.

> [card] [Test changes with evals](https://platform.atlan.com/evals/index.md)
>
> Score fixed cases before you release a change.

> [card] [Score production sessions](https://platform.atlan.com/evals/online)
>
> Judge real sessions against your rubric.

> [card] [Track spend and tokens](https://platform.atlan.com/finops/index.md)
>
> Compare cost across agents and check the rates behind it.

## Choose where to start

* **New to tracing?** [Trace your first agent run](https://platform.atlan.com/tools/sdk/tutorials/quickstart.md) records one run with the atlanai SDK and finds it under your agent.
* **Why did this run fail?** Start from the agent's **Sessions** or **Usage** tab, then [Inspect recorded run](https://platform.atlan.com/registry/traces/how-tos/inspect-runs.md).
* **Did a change improve the answers?** Start from a fixed dataset and an experiment, then [Review experiment results](https://platform.atlan.com/evals/how-tos/review-experiments.md).
* **Are real sessions meeting the rubric?** Start from a hosted scorer, then [Score production sessions](https://platform.atlan.com/evals/online).
* **Which agent drove the cost increase?** Start from **Reports → Agent usage**, then [Investigate agent cost](https://platform.atlan.com/finops/how-tos/investigate-cost.md).
* **Why is a skill's usage empty?** Check the recorded skill attribution in [How skill usage works](https://platform.atlan.com/registry/skills/concepts/usage.md).
* **Run missing or under the wrong identity?** [Trace errors](https://platform.atlan.com/registry/traces/troubleshooting.md).

> [!NOTE]
>
> Monitoring covers only the runs you send to Atlan Trident. Missing spans, scores, or costs do not show that a run succeeded or cost nothing. A trace error is an execution failure; an answer with no errors can still fail your rubric, so read traces and eval scores together. Run costs are calculated from traces and your rates; they are not your provider's invoice.

## See also

* [Trace attributes](https://platform.atlan.com/registry/traces/references/attributes.md): The OpenTelemetry attributes that set agent, session, model, tokens, and cost.
* [Trace limits and privacy](https://platform.atlan.com/registry/traces/references/limits-and-privacy.md): Ingest limits, the content policy, and content deletion.
* [Eval options and errors](https://platform.atlan.com/evals/reference.md): Every `Eval` option, status, limit, and error code.
* [Usage and cost measures](https://platform.atlan.com/finops/references/measures.md): What each report measure means.
* [Public API](https://platform.atlan.com/api/references/api-reference.md): Trace, eval, and OTLP ingest operations.


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