Call AI models across providers with Vercel AI Gateway
import { generateText } from 'ai';
const { text } = await generateText({
model: 'openai/gpt-6-astra',
prompt: 'Explain AI Gateway in one sentence.',
});
console.log(text);curl https://ai-gateway.vercel.sh/v1/chat/completions \
-H "Authorization: Bearer $AI_GATEWAY_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "openai/gpt-6-astra",
"messages": [
{
"role": "user",
"content": "Explain AI Gateway in one sentence."
}
]
}'AI Gateway is available on all plans. Your use of each AI provider is subject to their terms listed on each model's page and subject to Vercel's AI Product Terms.
Vercel AI Gateway gives applications and coding agents shared access to models across providers. It supports modalities including text generation, image generation, video generation, speech, transcription, realtime, embeddings, reranking, and decisions. Add files and tool calls where the selected model supports them.
Call AI Gateway from any environment with the AI SDK, OpenAI Chat Completions, OpenAI Responses API, Anthropic Messages, a supported coding agent, or a framework integration.
AI Gateway routes requests across providers and fallback models, then records status, provider, latency, token usage, cost, and every routing attempt. To prefer an inference provider and verify which provider served the request, set a provider order and read the response metadata. Configure budgets, access policies, data handling, and Bring Your Own Key (BYOK) for your team. AI Gateway adds zero markup to provider token prices, including with BYOK.
AI Gateway is a hosted API at https://ai-gateway.vercel.sh. Your application does not need to run on Vercel. Call AI Gateway from any server, cloud, or local environment with an AI Gateway API key. You need a Vercel account to create the key and manage billing, but not a Vercel project or deployment. Vercel deployments can use OpenID Connect (OIDC) instead of a key.
| Language | Connect with |
|---|---|
| TypeScript and JavaScript | The AI SDK, or the OpenAI or Anthropic SDK with the AI Gateway base URL |
| Python | The OpenAI or Anthropic Python SDK, the AI SDK for Python, or a framework integration such as LangChain, LiteLLM, or Pydantic AI |
| Go | The OpenAI or Anthropic Go SDK with the AI Gateway base URL |
| Other languages | Any client for the OpenAI Chat Completions, OpenAI Responses, or Anthropic Messages API, or the REST API |
Use AI Gateway when you want these controls without operating your own proxy, database, and routing control plane:
| Requirement | How AI Gateway handles it |
|---|---|
| Centralized credentials | Authenticate your application once, then use AI Gateway system credentials or your own provider accounts through BYOK. |
| Request logging | Inspect each request's model, provider attempts, latency, token usage, status, and cost in request logs. |
| Cost tracking | See spend by model in the dashboard and the cost of each request in request logs. Query spend by user, tag, or provider with Custom Reporting, which has its own pricing. |
| Spending controls | Set budgets for a team, project, API key, or team member. AI Gateway rejects new requests after an applicable budget is exceeded. |
| Caching | Set caching: 'auto' for prompt caching across providers. You pay the provider's reduced cache-read rate for cached input tokens. To reuse full responses, cache them in your application. |
| Provider failover | Route a model across healthy providers and configure ordered provider and model fallbacks. |
Budgets cover spend billed through AI Gateway system credentials. BYOK spend is metered separately and does not count toward those limits. If a BYOK request fails, AI Gateway can fall back to system credentials. After a budget is exceeded, AI Gateway rejects every request in that budget's scope with an HTTP 402 response. This includes BYOK requests. Review the BYOK behavior and soft-cap budget semantics when strict provider-account use or zero-overshoot limits are requirements.
Make your first request
Start with a coding agent, cURL, TypeScript, or Python.
Build an AI agent with AI SDK
Follow a step-by-step guide to add streaming, tools, and model fallbacks.
Use an SDK or API
Connect an existing AI SDK, OpenAI, Anthropic, or HTTP client.
Connect a coding agent
Route supported coding agents through AI Gateway with Vercel CLI.
Migrate existing model calls
Keep your current request format while moving model execution to AI Gateway.
Models and providers
Choose models and configure routing, filtering, fallbacks, caching, and service tiers.
Modalities
Work with text, images, video, speech, realtime, embeddings, reranking, and decisions.
Observability and spend
Inspect requests, usage, latency, routing attempts, and cost.
Authentication and BYOK
Authenticate with team-scoped API keys or OIDC, and connect provider credentials.
Security and compliance
Configure access, data handling, regional inference, and safety controls.
Pricing
Review model prices, credits, payment options, and discounts.
Explore chat platforms and ecosystem integrations, learn how to handle 429 responses, or see the AI Gateway FAQ for common questions about model availability, pricing, and request errors.
- Deploy a Shopify storefront with an AI assistant: Deploy an eve shopping agent with Vercel Shop to find Shopify products, manage carts, and answer store questions.
- Build AI agents in Python with the AI SDK: Stream model responses, get typed output, and build tool-using agents in Python on Vercel.
- How to control AI provider data handling with AI Gateway: Prevent AI providers from training on your prompts, enforce zero data retention, and restrict providers and models with Vercel AI Gateway.
- Running Durable Agents on Vercel: Design durable AI agents on Vercel with patterns for execution recovery, human approval, safe retries, deployment changes, and persistent files.
Was this helpful?