Tools
AI Gateway
Vercel’s unified API for hundreds of models — one endpoint, auth, failover, spend controls; pairs with the AI SDK and powers this wiki’s Ask models.
What it does
AI Gateway is Vercel’s model router: call many providers through one API with shared auth, usage tracking, failover/load balancing, and spend controls. It is designed to work tightly with the AI SDK while remaining usable from other OpenAI-compatible clients.
Key capabilities
- Single endpoint / model string routing across major providers
- Failover and capacity-oriented routing
- Usage metadata, routing logs, and spend monitoring
- Streaming, tools, and structured outputs via AI SDK integration
- Bring-your-own-key and enterprise controls (see current docs)
Data retention
The gateway deletes prompts and responses after completion by default. Upstream retention is separately controlled through provider agreements and opt-in ZDR routing. Security documentation
ZDR includes no prompt training; team-wide enforcement prevents callers from weakening the policy. Vercel explanation
Provider model/function exclusions and caching policies still matter. See vercel-ai-gateway-zero-data-retention for the distinction between content retention, operational records and application storage.
Canonical sources
- Docs (preferred): vercel.com/docs/ai-gateway
- Launch write-up: introducing-the-ai-gateway (
https://vercel.com/blog/ai-gateway) — historical; prefer docs for current behaviour - Models catalogue: vercel.com/ai-gateway/models
How this wiki uses it
Public Ask routes completions through Gateway (see Eve providerOptions.gateway and rerank via ai-gateway.vercel.sh). Prefer this hub over model-availability changelog leaves when answering “what is AI Gateway?”