# Analytics & Usage

Inspect request volume, errors, bytes, latency, credits, and AI attribution for each Ferrite app.

Ferrite folds Gateway traffic into per-app request, error, byte, average-latency,
and maximum-latency totals. Account statements keep funded credit and charges
separate from application telemetry.

## Read usage

Open **Usage** in the Dashboard for the account aggregate, then select the app
to correlate usage with its release and logs. MCP clients can call `usage` with
the app name for the same bounded record.

## AI attribution

The `ai-budgets` package adds tenant, application, model, provider, and purpose
dimensions to model calls. Keep provider billing evidence and app request
telemetry as related but distinct sources.

## Use metrics to answer a question

Start with a concrete question such as “Did error rate increase after release
42?” or “Which tenant exhausted its AI budget?” Narrow the time range and app
generation before reading individual logs.

See [Logs](/developers/logs/) and [AI Gateway](/developers/ai-gateway/).
