AI Gateway API key environment variable
The AI Gateway API key is configured via the environment variable AI_GATEWAY_API_KEY.
AI SDK · Providers · all subjects
ai gateway/setup
7 notes, read out of this brain and free to use. Each one was extracted from a source and is re-checked against its exam.
Importing gateway provider from AI SDK
The gateway provider can be imported in two ways: Option 1 imports from the 'ai' package (included by default). Option 2 installs and imports from '@ai-sdk/gateway' package separately. Both approaches allow calling gateway() with a model string like 'anthropic/claude-sonnet-4.5'.
AI Gateway provider is built into AI SDK
The Vercel AI Gateway provider is part of the AI SDK and does not require installation of additional provider modules or dependencies.
AI Gateway basic usage with plain model string
You can use AI Gateway directly by passing a model string in the format 'creator/model-name' to generateText, streamText, and other AI SDK functions. The AI SDK automatically uses AI Gateway when you pass a model string in this format. Example: const { text } = await generateText({ model: 'openai/gpt-5.4', prompt: 'Hello world' });
AI Gateway provider instance creation
Import the `gateway` provider instance from the 'ai' package (requires version 5.0.36 or later). You can also create a custom provider instance using `createGateway` from 'ai'.
AI Gateway custom provider instance options
When creating a custom AI Gateway provider instance with createGateway(), you can configure: baseURL (string, default: 'https://ai-gateway.vercel.sh/v4/ai'), apiKey (string, defaults to AI_GATEWAY_API_KEY env var, supports AI Gateway API keys and Vercel access tokens), teamIdOrSlug (string for multi-team Vercel access tokens), headers (Record<string,string> for custom headers), fetch (custom fetch implementation), metadataCacheRefreshMillis (number, defaults to 300,000ms).
AI Gateway Bring Your Own Key (BYOK) credentials
You can connect your own provider credentials to use with Vercel AI Gateway by adding provider credentials in your Vercel team's AI Gateway settings. No code changes are needed. For providers like Azure that support custom deployment names, you can configure model mappings to map gateway model slugs to your deployment names.