> ## 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 AI Registry, Atlan's hosted product. The AI registry specification and aireg CLI are maintained separately. /open-source lists publicly reachable signed installers and explains that the source repository and hosted CLI reference currently require GitHub access. Do not claim the repository or its reference docs are publicly accessible. Use these docs for Atlan-specific setup such as the adapter, sign-in and workspaces.

# Usage and cost measures

> Interpret spend, run counts, token categories, cache usage, error rates, and latency in agent reports.

Agent reports summarize recorded execution data for the selected scope and period. Keep the workspace and time range the same when comparing a report with an agent profile.

| Measure | Meaning | How to read it |
| - | - | - |
| **Spend** | Sum of recorded costs in the period | Check the rates and price sources before comparing it with a provider invoice |
| **Runs** | Recorded agent runs in the period | Missing telemetry means the report can undercount actual executions |
| **Active agents** | Agents with recorded activity in the period | A registered agent with no recorded runs is not active in this measure |
| **Cost / run** | Period spend divided by recorded runs | Read alongside run count and token usage |
| **Tokens / run** | Recorded tokens divided by runs | Averages can conceal a small number of large runs |
| **Span error rate** | Share of recorded spans marked as errors | Measures execution errors; use eval scores for answer quality |
| **P90 latency** | The 90th percentile of the measured run durations | Use the timeline to identify the slow call or extra steps |
| **Runs using skills** | Runs with recorded skill-use evidence | Missing attribution does not establish that no skill was used |

## Token categories

Atlan prices fresh input, output, cache reads, and cache writes separately. Provider counts can overlap: a provider may include cached tokens in its input total. The trace projection normalizes supported token fields into separate buckets before pricing, so cached input is not charged again as fresh input.

Reasoning tokens can be a subset of output tokens. Do not add every token attribute together to reconstruct a bill; use the report's measures and the recorded provider semantics.

## Missing and unpriced values

An unavailable breakdown is different from a zero. **Unpriced** means pricing found no applicable model rate. A zero beside an unpriced run does not establish that the provider charged nothing.

When the rate service cannot be reached, telemetry can retain the cost sent by its producer. A record with missing model or token information cannot be priced from token rates. See [Run pricing](/finops/index#run-pricing).

## Reporting boundaries

Reports cover telemetry that reached Atlan and that the viewer can access. They do not reconcile provider invoices, account for unrecorded calls, or enforce spending limits. The [Models & rates](/admin/finops/models-and-rates) settings determine pricing; [evals](/evals/index) determine whether the measured behavior meets your criteria.


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