Skip to content
Dashboard

What is an AI gateway?

Copy link to headingWhat is an AI gateway?

Copy link to headingHow an AI gateway differs from a router or a traditional API gateway

Copy link to headingSeven failure modes an AI gateway prevents

Copy link to headingHow AI Gateway routes, fails over, and reports on every request

Copy link to headingThe infrastructure runs on Vercel's existing network

Copy link to headingRouting, fallbacks, and per-provider timeouts

import { streamText } from 'ai';
const result = streamText({
model: 'anthropic/claude-opus-4.8', // defaults to Vercel AI Gateway
prompt: 'How does AI Gateway route around a provider outage?',
});

Copy link to headingKeys stay off the application and off the request path

Copy link to headingCost and performance show up per request, not per invoice

Copy link to headingDoes an AI gateway add latency?

Copy link to headingWhen do teams need a multi-model AI gateway?

Copy link to headingHow AI Gateway helps engineering teams ship reliably

Copy link to headingStop rewriting adapters every time a new model ships

Copy link to headingShip customer-requested features without touching keys

Copy link to headingMove retries and fallback logic out of application code

Copy link to headingSkip the build-versus-buy math on gateway infrastructure

Copy link to headingShip your AI Gateway integration

Copy link to headingFAQs about AI gateways

Copy link to headingWhat's the difference between AI Gateway and the AI SDK?

Copy link to headingDoes adding an AI gateway slow down LLM calls?

Copy link to headingCan I use an AI gateway with providers outside its catalog?

Copy link to headingDoes Vercel store or train on prompts that pass through AI Gateway?

Ready to deploy?