AI Gateway Overview

Use popular AI models in your code, without needing to manage API keys or external accounts.

This feature is available on Credit-based plans only, including the Free, Personal, and Pro plans. If you are on an Enterprise plan and you’re interested, reach out to your Account Manager.

Overview

The AI Gateway service simplifies technical and operational concerns when using AI inference in your code, by removing the need to:

• Open an account with each provider you want to use.
• Maintain a separate credit balance with each provider.
• Copy the API key from each provider to your projects on Netlify.

Explore an AI Gateway example

Learn how AI Gateway works in this TanStack Start chat app example, which automatically proxies your requests to a supported AI provider with built-in security, usage analytics, and rate limiting.

• Demo repo
• Demo site
• How AI Gateway works in this example

AI Gateway examples

For a video overview of how the AI Gateway works with a fun demo project, check out our AI Gateway gameshow demo.

Check out more examples of working projects that are powered with AI models in our AI Gateway examples docs.

How it works

By default, Netlify automatically sets the appropriate environment variables that AI client libraries typically use for configuration, in all Netlify compute contexts (e.g., Netlify Functions, Edge Functions, Preview Server, etc.).

These variables include:

• API keys for OpenAI, Anthropic, Google Gemini, and OpenRouter.
• A custom base URL for each provider, to route requests via the AI Gateway service.

These variables are picked up by the official client libraries of these providers, so no extra configuration is necessary - with the exception of the OpenRouter SDK, which needs the base URL passed explicitly. Alternatively, if you make AI calls via a provider's REST API, these values are easy to incorporate in your code.

When receiving a request from a client, the AI Gateway makes the call to the AI provider on your behalf. Then, it bills your Netlify account by converting the actual token usage in the request into credits, using your existing credit quota.

Local development with the AI Gateway

The AI Gateway has full support with the Netlify CLI. For Vite-based projects, you can also use the Netlify Vite plugin to access AI Gateway locally without running netlify dev. Check our Quickstart for a hands-on guided example project using the AI Gateway.

Note: A project must have a production deploy for the AI Gateway to activate, so if you're creating a new project locally, deploy to production at least once to enable it.

The AI Gateway does not store your prompts or model outputs. Learn more about Security and Privacy for AI features. To opt out, check your opt-out options.

Support in web frameworks

When you develop server-side code with any web framework supported by Netlify (e.g., Astro; Tanstack Start; Next.js, Gatsby, Nuxt, etc.), your code is packaged in Netlify Functions and Edge Functions under the hood, as part of the build process.

Therefore, the above environment variables are available as when explicitly using Netlify compute primitives, without any further settings required.

Using the AI Gateway

For a quickstart, check out our Quickstart for AI Gateway.

The AI Gateway is available by default in all credit-based plans, unless:

1. You have disabled Netlify AI Features for your team, or:
2. You have set your own API keys for AI providers via environment variables. Netlify does not override these keys. You can add or remove your own keys at any point.

For full information on which environment variables are automatically set, and how to control this behavior, see here.

Using official client libraries

If you're using any of the following libraries, little to no configuration is required. The AI Gateway automatically provides the necessary environment variables that these libraries use:

Anthropic Claude

import Anthropic from '@anthropic-ai/sdk';

const anthropic = new Anthropic();
// No API key or base URL configuration needed - automatically uses:
// process.env.ANTHROPIC_API_KEY and process.env.ANTHROPIC_BASE_URL

async function callAnthropic() {
  const message = await anthropic.messages.create({
    model: 'claude-sonnet-4-5-20250929',
    max_tokens: 1024,
    messages: [{ role: 'user', content: 'Hello!' }]
  });
  return message;
}

Library: Anthropic TypeScript API Library

OpenAI

import OpenAI from 'openai';

const openai = new OpenAI();
// No API key or base URL configuration needed - automatically uses:
// process.env.OPENAI_API_KEY and process.env.OPENAI_BASE_URL

async function callOpenAI() {
  const completion = await openai.chat.completions.create({
    // Models available through OpenRouter can also be used here, with no extra
    // configuration - just pass a model name in the OpenRouter notation,
    // e.g. 'deepseek/deepseek-v4-flash-0731'
    model: 'gpt-5',
    messages: [{ role: 'user', content: 'Hello!' }]
  });
  return completion;
}

Library: OpenAI TypeScript and JavaScript API Library

Google Gemini

import { GoogleGenAI } from '@google/genai';

const genAI = new GoogleGenAI({});
// No API key or base URL configuration needed - automatically uses:
// process.env.GEMINI_API_KEY and process.env.GOOGLE_GEMINI_BASE_URL

async function callGemini() {
  const result = await genAI.models.generateContent({
    model: 'gemini-2.5-pro',
    contents: 'Hello!'
  });
  return result;
}

Library: Google Gen AI SDK for TypeScript and JavaScript

OpenRouter

import { OpenRouter } from '@openrouter/sdk';

const openRouter = new OpenRouter();
// No API key or base URL configuration needed - automatically uses:
// process.env.OPENROUTER_API_KEY and process.env.OPENROUTER_BASE_URL
// Requires @openrouter/sdk 1.2.43 or later. Earlier versions ignore
// OPENROUTER_BASE_URL and call openrouter.ai directly, which rejects the
// Netlify-issued key with "401 Missing Authentication header".

async function callOpenRouter() {
  const result = await openRouter.chat.send({
    chatRequest: {
      model: 'x-ai/grok-4.5',
      messages: [{ role: 'user', content: 'Hello!' }]
    }
  });
  return result;
}

Library: OpenRouter TypeScript SDK

Note that models available through OpenRouter can be used with either the OpenRouter SDK or the OpenAI SDK.

Using official REST APIs

Anthropic Claude

async function callAnthropic() {
  const ANTHROPIC_API_KEY = process.env.ANTHROPIC_API_KEY;
  const ANTHROPIC_BASE_URL = process.env.ANTHROPIC_BASE_URL;

  const response = await fetch(`${ANTHROPIC_BASE_URL}/v1/messages`, {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'x-api-key': ANTHROPIC_API_KEY,
      'anthropic-version': '2023-06-01'
    },
    body: JSON.stringify({
      model: 'claude-sonnet-4-5-20250929',
      max_tokens: 1024,
      messages: [{ role: 'user', content: 'Hello!' }]
    })
  });
  return await response.json();
}

OpenAI

async function callOpenAI() {
  const OPENAI_API_KEY = process.env.OPENAI_API_KEY;
  const OPENAI_BASE_URL = process.env.OPENAI_BASE_URL;

  const response = await fetch(`${OPENAI_BASE_URL}/v1/chat/completions`, {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'Authorization': `Bearer ${OPENAI_API_KEY}`
    },
    body: JSON.stringify({
      model: 'gpt-5',
      messages: [{ role: 'user', content: 'Hello!' }]
    })
  });
  return await response.json();
}

Google Gemini

async function callGemini() {
  const GEMINI_API_KEY = process.env.GEMINI_API_KEY;
  const GEMINI_BASE_URL = process.env.GOOGLE_GEMINI_BASE_URL;

  const response = await fetch(
    `${GEMINI_BASE_URL}/v1beta/models/gemini-2.5-pro:generateContent`,
    {
      method: 'POST',
      headers: {
        'Content-Type': 'application/json',
        'x-goog-api-key': GEMINI_API_KEY
      },
      body: JSON.stringify({
        contents: [{
          parts: [{ text: 'Hello!' }]
        }]
      })
    }
  );
  return await response.json();
}

OpenRouter

async function callOpenRouter() {
  const OPENROUTER_API_KEY = process.env.OPENROUTER_API_KEY;
  const OPENROUTER_BASE_URL = process.env.OPENROUTER_BASE_URL;

  const response = await fetch(`${OPENROUTER_BASE_URL}/chat/completions`, {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'Authorization': `Bearer ${OPENROUTER_API_KEY}`
    },
    body: JSON.stringify({
      model: 'x-ai/grok-4.5',
      messages: [{ role: 'user', content: 'Hello!' }]
    })
  });
  return await response.json();
}

Using third-party client libraries

If you are using a client library that does not work out-of-the-box with the environment variables set for the AI Gateway, you need to manually pass the API key and base URL as arguments to the library.

This is similar to manually reading & passing variable values when using a provider's REST API. See Using official REST APIs above for the relevant variable names.

Managing environment variables

If you have already set an API key or base URL at the project or team level, Netlify will never override it.

When a Netlify Function or Edge Function is initialized, the following environment variables are set to the appropriate values for the AI Gateway:

1. OPENAI_API_KEY and OPENAI_BASE_URL - unless any of these is already set by you at the project or team level.
2. ANTHROPIC_API_KEY and ANTHROPIC_BASE_URL - unless any of these is already set by you.
3. GEMINI_API_KEY and GOOGLE_GEMINI_BASE_URL- unless any of these is already set by you, or if either GOOGLE_API_KEY or GOOGLE_VERTEX_BASE_URL are set.
4. OPENROUTER_API_KEY and OPENROUTER_BASE_URL - unless any of these is already set by you.

NETLIFY_AI_GATEWAY_KEY and NETLIFY_AI_GATEWAY_BASE_URL environment variables are always injected into the AI Gateway-supported runtimes. If you want to mix different setups with your own keys and Netlify's or you want to be explicit about using AI Gateway keys in your calls, use these env variables as they will never collide with other environment variables values.

To prevent any variables from being automatically set, you can disable AI Features.

Model availability

Models are available through the AI Gateway in two ways.

Models served directly

Models from Anthropic, OpenAI, and Google Gemini are served directly by the AI Gateway, using each provider's own API. These are the models listed below, and they are not routed through OpenRouter.

AI Provider	Model
Anthropic	claude-fable-5
Anthropic	claude-haiku-4-5
Anthropic	claude-haiku-4-5-20251001
Anthropic	claude-opus-4-5
Anthropic	claude-opus-4-5-20251101
Anthropic	claude-opus-4-6
Anthropic	claude-opus-4-7
Anthropic	claude-opus-4-8
Anthropic	claude-opus-5
Anthropic	claude-sonnet-4-5
Anthropic	claude-sonnet-4-5-20250929
Anthropic	claude-sonnet-4-6
Anthropic	claude-sonnet-5
Gemini	gemini-2.5-flash
Gemini	gemini-2.5-flash-image
Gemini	gemini-2.5-flash-lite
Gemini	gemini-2.5-pro
Gemini	gemini-3-flash-preview
Gemini	gemini-3-pro-image
Gemini	gemini-3.1-flash-image
Gemini	gemini-3.1-flash-lite
Gemini	gemini-3.1-flash-lite-image
Gemini	gemini-3.1-pro-preview
Gemini	gemini-3.1-pro-preview-customtools
Gemini	gemini-3.5-flash
Gemini	gemini-3.5-flash-lite
Gemini	gemini-3.6-flash
Gemini	gemini-3.7-flash
Gemini	gemini-flash-latest
Gemini	gemini-flash-lite-latest
OpenAI	chat-latest
OpenAI	gpt-4.1
OpenAI	gpt-4.1-mini
OpenAI	gpt-4.1-nano
OpenAI	gpt-4o
OpenAI	gpt-4o-mini
OpenAI	gpt-5
OpenAI	gpt-5-2025-08-07
OpenAI	gpt-5-mini
OpenAI	gpt-5-mini-2025-08-07
OpenAI	gpt-5-nano
OpenAI	gpt-5-pro
OpenAI	gpt-5.1
OpenAI	gpt-5.1-2025-11-13
OpenAI	gpt-5.2
OpenAI	gpt-5.2-2025-12-11
OpenAI	gpt-5.2-pro
OpenAI	gpt-5.2-pro-2025-12-11
OpenAI	gpt-5.3-codex
OpenAI	gpt-5.4
OpenAI	gpt-5.4-2026-03-05
OpenAI	gpt-5.4-mini
OpenAI	gpt-5.4-mini-2026-03-17
OpenAI	gpt-5.4-nano
OpenAI	gpt-5.4-nano-2026-03-17
OpenAI	gpt-5.4-pro
OpenAI	gpt-5.4-pro-2026-03-05
OpenAI	gpt-5.5
OpenAI	gpt-5.5-2026-04-23
OpenAI	gpt-5.5-pro
OpenAI	gpt-5.5-pro-2026-04-23
OpenAI	gpt-5.6-luna
OpenAI	gpt-5.6-sol
OpenAI	gpt-5.6-terra
OpenAI	o3
OpenAI	o3-mini
OpenAI	o4-mini
Openrouter	allenai/olmo-3-32b-think
Openrouter	anthracite-org/magnum-v4-72b
Openrouter	arcee-ai/trinity-large-thinking
Openrouter	arcee-ai/virtuoso-large
Openrouter	baidu/ernie-4.5-vl-424b-a47b
Openrouter	bytedance-seed/seed-1.6
Openrouter	bytedance-seed/seed-1.6-flash
Openrouter	bytedance-seed/seed-2-1-turbo
Openrouter	bytedance-seed/seed-2.0-code
Openrouter	bytedance-seed/seed-2.0-lite
Openrouter	bytedance-seed/seed-2.0-mini
Openrouter	bytedance/ui-tars-1.5-7b
Openrouter	cognitivecomputations/dolphin-mistral-24b-venice-edition
Openrouter	deepcogito/cogito-v2.1-671b
Openrouter	deepseek/deepseek-chat
Openrouter	deepseek/deepseek-chat-v3-0324
Openrouter	deepseek/deepseek-chat-v3.1
Openrouter	deepseek/deepseek-r1
Openrouter	deepseek/deepseek-r1-0528
Openrouter	deepseek/deepseek-r1-distill-llama-70b
Openrouter	deepseek/deepseek-v3.1-terminus
Openrouter	deepseek/deepseek-v3.2
Openrouter	deepseek/deepseek-v3.2-exp
Openrouter	deepseek/deepseek-v4-flash
Openrouter	deepseek/deepseek-v4-flash-0731
Openrouter	deepseek/deepseek-v4-pro
Openrouter	deepseek/deepseek-v4-pro-0813
Openrouter	google/gemma-2-27b-it
Openrouter	google/gemma-3-12b-it
Openrouter	google/gemma-3-27b-it
Openrouter	google/gemma-3-4b-it
Openrouter	google/gemma-3n-e4b-it
Openrouter	google/gemma-4-26b-a4b-it
Openrouter	google/gemma-4-31b-it
Openrouter	gryphe/mythomax-l2-13b
Openrouter	ibm-granite/granite-4.1-8b
Openrouter	inception/mercury-2
Openrouter	inclusionai/ling-2.6-1t
Openrouter	inclusionai/ling-2.6-flash
Openrouter	inclusionai/ling-3.0-flash
Openrouter	inclusionai/ring-2.6-1t
Openrouter	mancer/weaver
Openrouter	meta-llama/llama-3.1-70b-instruct
Openrouter	meta-llama/llama-3.1-8b-instruct
Openrouter	meta-llama/llama-3.2-3b-instruct
Openrouter	meta-llama/llama-3.3-70b-instruct
Openrouter	meta-llama/llama-4-maverick
Openrouter	meta-llama/llama-4-scout
Openrouter	meta-llama/llama-guard-4-12b
Openrouter	meta/muse-glimmer-30b
Openrouter	microsoft/phi-4
Openrouter	microsoft/wizardlm-2-8x22b
Openrouter	minimax/minimax-m1
Openrouter	minimax/minimax-m2
Openrouter	minimax/minimax-m2.1
Openrouter	minimax/minimax-m2.5
Openrouter	minimax/minimax-m2.7
Openrouter	minimax/minimax-m3
Openrouter	mistralai/ministral-8b
Openrouter	mistralai/mistral-nemo
Openrouter	mistralai/mistral-small-24b-instruct-2501
Openrouter	mistralai/mistral-small-2603
Openrouter	mistralai/mistral-small-3.2-24b-instruct
Openrouter	moonshotai/kimi-k2
Openrouter	moonshotai/kimi-k2-0905
Openrouter	moonshotai/kimi-k2-thinking
Openrouter	moonshotai/kimi-k2.5
Openrouter	moonshotai/kimi-k2.6
Openrouter	moonshotai/kimi-k2.7-code
Openrouter	moonshotai/kimi-k3
Openrouter	morph/morph-v3-fast
Openrouter	morph/morph-v3-large
Openrouter	nousresearch/hermes-3-llama-3.1-405b
Openrouter	nousresearch/hermes-3-llama-3.1-70b
Openrouter	nousresearch/hermes-4-405b
Openrouter	nousresearch/hermes-4-70b
Openrouter	nvidia/nemotron-3-nano-30b-a3b
Openrouter	nvidia/nemotron-3-super-120b-a12b
Openrouter	nvidia/nemotron-3-ultra-550b-a55b
Openrouter	nvidia/nemotron-3.5-lightning
Openrouter	openai/gpt-oss-120b
Openrouter	openai/gpt-oss-20b
Openrouter	openai/gpt-oss-safeguard-20b
Openrouter	openrouter/auto
Openrouter	openrouter/auto-beta
Openrouter	openrouter/bodybuilder
Openrouter	openrouter/free
Openrouter	openrouter/fusion
Openrouter	openrouter/pareto-code
Openrouter	perceptron/perceptron-mk1
Openrouter	perplexity/sonar
Openrouter	perplexity/sonar-deep-research
Openrouter	perplexity/sonar-pro
Openrouter	perplexity/sonar-pro-search
Openrouter	perplexity/sonar-reasoning-pro
Openrouter	qwen/qwen-2.5-72b-instruct
Openrouter	qwen/qwen-2.5-7b-instruct
Openrouter	qwen/qwen2.5-vl-72b-instruct
Openrouter	qwen/qwen3-14b
Openrouter	qwen/qwen3-235b-a22b-2507
Openrouter	qwen/qwen3-235b-a22b-thinking-2507
Openrouter	qwen/qwen3-30b-a3b
Openrouter	qwen/qwen3-30b-a3b-instruct-2507
Openrouter	qwen/qwen3-32b
Openrouter	qwen/qwen3-coder
Openrouter	qwen/qwen3-coder-30b-a3b-instruct
Openrouter	qwen/qwen3-coder-next
Openrouter	qwen/qwen3-next-80b-a3b-instruct
Openrouter	qwen/qwen3-next-80b-a3b-thinking
Openrouter	qwen/qwen3-vl-235b-a22b-instruct
Openrouter	qwen/qwen3-vl-235b-a22b-thinking
Openrouter	qwen/qwen3-vl-30b-a3b-instruct
Openrouter	qwen/qwen3-vl-8b-instruct
Openrouter	qwen/qwen3.5-122b-a10b
Openrouter	qwen/qwen3.5-27b
Openrouter	qwen/qwen3.5-35b-a3b
Openrouter	qwen/qwen3.5-397b-a17b
Openrouter	qwen/qwen3.5-9b
Openrouter	qwen/qwen3.6-27b
Openrouter	qwen/qwen3.6-35b-a3b
Openrouter	qwen/qwen3.8-2.4t-a95b
Openrouter	qwen/qwen3.8-27b
Openrouter	rekaai/reka-edge
Openrouter	rekaai/reka-flash-3
Openrouter	relace/relace-apply-3
Openrouter	relace/relace-search
Openrouter	sao10k/l3-lunaris-8b
Openrouter	sao10k/l3.1-euryale-70b
Openrouter	sao10k/l3.3-euryale-70b
Openrouter	stepfun/step-3.7-flash
Openrouter	tencent/hy-mt2-1.8b
Openrouter	tencent/hy-mt2-30b-a3b
Openrouter	tencent/hy3
Openrouter	thedrummer/cydonia-24b-v4.1
Openrouter	thedrummer/rocinante-12b
Openrouter	thedrummer/skyfall-36b-v2
Openrouter	thedrummer/unslopnemo-12b
Openrouter	thinkingmachines/inkling
Openrouter	thinkingmachines/inkling-small
Openrouter	undi95/remm-slerp-l2-13b
Openrouter	x-ai/grok-4.20
Openrouter	x-ai/grok-4.20-multi-agent
Openrouter	x-ai/grok-4.3
Openrouter	x-ai/grok-4.5
Openrouter	x-ai/grok-4.6
Openrouter	x-ai/grok-build-0.1
Openrouter	xiaomi/mimo-v2.5
Openrouter	xiaomi/mimo-v2.5-pro
Openrouter	z-ai/glm-4.5-air
Openrouter	z-ai/glm-4.5v
Openrouter	z-ai/glm-4.6
Openrouter	z-ai/glm-4.6v
Openrouter	z-ai/glm-4.7
Openrouter	z-ai/glm-4.7-flash
Openrouter	z-ai/glm-5
Openrouter	z-ai/glm-5.1
Openrouter	z-ai/glm-5.2
Openrouter	z-ai/glm-5.2:free
Openrouter	~deepseek/deepseek-v4-flash-latest
Openrouter	~moonshotai/kimi-latest
Openrouter	~x-ai/grok-latest

Models served via OpenRouter

Models from other model creators, such as xAI, DeepSeek, Meta, Mistral, and Qwen, are available in partnership with OpenRouter, which routes many AI models. To browse the full catalog and find the model IDs to use, see the OpenRouter models directory.

To call these models, pass a model ID in the OpenRouter notation - for example, deepseek/deepseek-v4-flash-0731 - to either the OpenRouter SDK or the OpenAI SDK, as shown in Using official client libraries above. You can also call them via REST, as shown in Using official REST APIs.

Zero Data Retention only

The same model on OpenRouter can be hosted by multiple providers (i.e., vendors that offer that model). Netlify only routes your requests to providers that have a Zero Data Retention (ZDR) policy, meaning they do not store your prompts or model outputs.

As a result, if none of the providers hosting a given model offers a ZDR guarantee, that model is not served by the AI Gateway - even though it is listed in the OpenRouter models directory.

Pricing

To understand pricing for AI Gateway, check out our Pricing for AI features docs.

Rate limits

Netlify applies rate limits for your team's usage of the AI Gateway, across all of your team's projects. The rate limit is per minute and differs by plan, with higher plans having a higher limit.

Tokens consumed by each request to the AI Gateway are converted to USD (U.S. dollars) based on the costs published by the providers we support, and then to Netlify credits. $1 USD of AI model usage equates to 180 credits. Learn more in our Pricing for AI features.

The limit depends on your plan:

Plan	Limit per minute (credits)
Free	90
Personal	450
Pro	1,800
Enterprise	9,000

Thus, when adding AI Gateway-based features to your site, it's always advised to track your credit usage - and ensure you have either enabled auto recharge or purchased credit packs to meet the demand you expect without exhausting your credits.

Configuring advanced rate limiting

We recommend that you set up rate limiting rules for Netlify functions or edge fuctions that use the AI gateway.

This lets you limit any given visitor from abusing calls to the AI Gateway, thus avoiding high costs over time or hitting your account-wide AI gateway limits.

Limitations

The AI Gateway has the following limitations at this time:

1. Using the AI Gateway requires that the site has had at least one production deployment in the past.
2. The context window (input prompt) is limited to 200k tokens.
3. Prompt caching:
   • Anthropic Claude: only the default 5-minute ephemeral cache duration is supported for Claude.
   • OpenAI: the AI Gateway sets a per-account prompt_cache_key.
   • Google Gemini: explicit context caching is not supported.
4. The AI Gateway does not pass through any request headers (and thus you cannot enable proprietary experimental features via headers).
5. Batch inference is not supported.
6. Priority processing (an OpenAI feature) is not supported.

Monitor AI Gateway usage

To help you monitor AI Gateway usage, check out our docs on monitoring AI feature usage.