Sign InOpen Brain
VercelEngineering PostOfficial Source

AI Gateway logs now have a dedicated page

AI Gateway’s dedicated logs expose per-request cost, tokens, latency, routing, and provider fallbacks, making agent failures and spend anomalies easier to trace.

Vercel · Jul 31, 2026
Open Source Open MarkdownOpen JSON
Source Summary

The new **Logs page** lists gateway requests with cost, token counts, duration, model, provider, and region. Request details include time to first token, cache and reasoning tokens, policy flags, and the ordered fallback path.

Practical Implication

Use filters and request-ID search to debug agent failures or compare routing behavior. Filtered views are shareable by URL and exportable as **CSV or JSON**, which makes targeted cost and reliability analysis easier.

Agent-Ready Context
The new **Logs page** lists gateway requests with cost, token counts, duration, model, provider, and region. Request details include time to first token, cache and reasoning tokens, policy flags, and the ordered fallback path.

Use filters and request-ID search to debug agent failures or compare routing behavior. Filtered views are shareable by URL and exportable as **CSV or JSON**, which makes targeted cost and reliability analysis easier.

The page records what happened but does not define alerts, retention, or automated remediation. Diagnosis still requires interpreting provider failures, routing-budget exhaustion, and latency in the context of the agent run.
Connected Context · Feed7 Judgment

This makes gateway behavior inspectable at request level by joining cost, latency, token, policy, provider, and fallback evidence in one exportable surface. It supports debugging and routing comparisons, but confirms that observability alone does not close the loop: teams still need alerts, retention choices, run-level interpretation, and remediation outside the page.

AI Gateway now supports team and project spend budgetsBudgets and alerts add warning and fail-closed controls to the spend evidence exposed here, while logs help explain which requests and scopes drove a cap to exhaustion.TokTier: Exact Stateful Tokenization for Agentic LLM ServingThe page's time-to-first-token and token details can help expose the latency class TokTier targets, though adopting stateful tokenization would still require more specific tokenization telemetry and correctness checks.Claude Fable 5 access restored on AI GatewayOrdered fallback paths and policy flags provide direct diagnostic evidence for the refusals and fallback configuration consequences described for Claude Fable 5.From Signal to PR: Anatomy of a Self-Improving Agent — Jason Lopatecki, ArizeThe Logs page supplies production trace evidence that a signal-to-PR agent could investigate, but the candidate's broader loop still requires evals, repository context, remediation logic, and human review.
Context Map
infracodingdata#observability#gateways#agent-reliability
Uncertainty
The page records what happened but does not define alerts, retention, or automated remediation. Diagnosis still requires interpreting provider failures, routing-budget exhaustion, and latency in the context of the agent run.