Export AI Gateway traces with Vercel Drains
Vercel AI Gateway now exports full request traces to any observability platform — $0.05 per 1K traces for Pro and Enterprise teams.
Why it matters
A practical feature for teams running AI workloads at scale on Vercel: native observability into latency, routing, token usage, and provider fallbacks without exposing prompt/completion data. Actionable for practitioners optimizing inference costs and reliability.
The key facts
14 to knowVercel AI Gateway traces now exportable via OpenTelemetry/OTLP-HTTP
Native integrations: Braintrust, Dash0, Kubiks, Sentry, Statsig
Pricing: $0.05 per 1,000 traces + $0.50/GB standard Drains rate
Traces include: model/provider routing, fallback/retry attempts, token usage, latency, response status, attribution tags
Traces exclude prompt and completion content for privacy
Sampling controls available for traffic filtering
Pro and Enterprise tier feature; failed deliveries not charged
OpenTelemetry traces for every AI Gateway request
Pricing: $0.05 per 1,000 traces + $0.50/GB data transfer (Drains rate)
Traces include: model/provider routing, fallback/retry attempts, token usage, TTFT, latency, status, attribution (project, deployment, API key, environment, custom tags)
Sampling controls available
Prompt/completion content excluded from traces
Pro and Enterprise teams only
Failed deliveries do not incur trace charges
Go to the source
Vercel Blogvercel.com
Publisher excerpt: AI Gateway now produces an OpenTelemetry trace for every request. Pro and Enterprise teams can send these traces through Vercel Drains to any OTLP/HTTP-compatible endpoint, including native integrations for Braintrust, Dash0, Kubiks, Sentry, and Statsig. Each trace shows the full request lifecycle,…