AI Gateway logs now have a dedicated page

Vercel has shipped a dedicated Logs page for AI Gateway. It lists every request a team sends through the gateway, newest first, and surfaces cost, token counts, duration, plus the model, provider, and region that served each call. The view is available at team scope and per project.

Open any request and you can see how it was routed. The page also supports filter and search: filter by provider, model, modality, credentials, or status so you can narrow a high-volume stream down to a specific path, credential, or failure mode instead of scanning a raw chronological dump.

For engineers building on multi-model stacks, that combination matters because gateway traffic often spans several providers and regions. Cost, tokens, and duration on the same row as model and provider make it practical to spot expensive routes, slow regions, or unexpected model choices without stitching logs from separate dashboards.

In market terms, AI Gateway sits in the layer that sits between apps and many model backends. A team-level and project-level log surface with routing detail is the operational counterpart to that routing layer: it turns “which provider answered, with what spend and latency” into a first-class product view rather than an afterthought in app-level telemetry.

If you already send traffic through AI Gateway, the practical next step is to open the Logs page at team and project scope, filter by the providers and models you care about, and inspect a few recent requests to confirm routing, cost, and token behavior match what you expect. Watch how status and credential filters behave under real load, and treat any mismatch between intended route and what the request detail shows as a signal to adjust gateway configuration.