AI Gateway adds a dedicated logs page

AI Gateway has added a dedicated Logs page where teams can view all requests sent through the gateway, sorted by most recent. It provides per-request cost, token counts, processing time, and the model, provider, and region used, and is available at both the team and project scope.

You can filter requests by provider, model, modality, credentials, and status, or search by model, provider, or request ID. Filter state is saved in the URL so it can be shared, and you can drag across a specific section of the request-count-over-time chart to inspect a detailed range, as well as export the current view as CSV or JSON.

Opening an individual request shows the following information.

• Total cost and a breakdown of input, output, reasoning, and cache read/write tokens
• Processing time and time to first token
• Whether Zero Data Retention and region restrictions were applied
• The provider fallback path attempted while processing the request

The fallback path shows the order, model, provider, status code, credentials, and processing time for each provider attempt. For failed attempts, the cause—such as a provider timeout or routing budget exhaustion—is also provided.