Getting Started with AI Gateway

Use AI Gateway from any environment to call models, connect coding agents, and inspect routing and spend. AI Gateway API keys belong to a Vercel team and work in local development, CI, external servers, and Vercel deployments.

Use this path when you want a coding agent to write and run an AI Gateway request in your current environment.

Paste this prompt into a coding agent with terminal access:

Use this path when you want to route a coding agent's own model requests through AI Gateway. The Vercel CLI detects supported agents, including Claude Code, and lets you select which ones to connect:

npx vercel@latest ai-gateway setup

The command provisions or reuses an API key, previews its configuration changes, and connects the agents you select. For Claude Code, it uses the dedicated AI Gateway endpoint and adds Gateway models to the /model picker. See Coding Agents for supported agents and manual setup.

You need a Vercel account. To use free AI Gateway Credits, add a valid payment method to your team. The cURL path needs no runtime. The TypeScript path needs Node.js 22.18 or later, and the Python path needs Python 3.12 or later with uv.

Create and export an API key:

1. Open the Create API Key dialog, enter a name, and select Create API Key. The link opens the dialog directly.
2. Copy the key immediately and export it in the terminal where you will run the example. You cannot retrieve its value again.

export AI_GATEWAY_API_KEY="your_ai_gateway_api_key"

These examples use openai/gpt-6-astra and consume AI Gateway Credits. You can replace the model slug with any model your team can access. If your team has not purchased credits, choose a free-tier model.

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": "Invent a new holiday and describe its traditions."
      }
    ]
  }'

pnpm add ai@latest

import { generateText } from 'ai';
 
const { text } = await generateText({
  model: 'openai/gpt-6-astra',
  prompt: 'Invent a new holiday and describe its traditions.',
});
 
console.log(text);

node index.mts

import asyncio
import ai
 
 
async def main() -> None:
    model = ai.get_model('openai/gpt-6-astra')
    messages = [
        ai.user_message('Invent a new holiday and describe its traditions.')
    ]
 
    async with ai.stream(model, messages) as stream:
        async for event in stream:
            if isinstance(event, ai.events.TextDelta):
                print(event.chunk, end='', flush=True)
 
    print()
 
 
if __name__ == '__main__':
    asyncio.run(main())

uv run --with ai quickstart.py

After the request completes, open AI Gateway Logs and select the newest request. You can inspect its status, model, provider, token usage, cost, duration, and routing attempts.

Logs may take up to 90 seconds to appear. Refresh the page if you don't see the request.

Use a supported SDK or API to connect an existing client. To move calls from a provider endpoint without changing their behavior, migrate to AI Gateway.

• Troubleshoot request errors by status code, type, and message
• Generate text, images, video, realtime audio, or speech
• Answer typed questions about application state in the Decision quickstart
• Browse models by capability, provider, price, and free-tier availability, or query GET /v1/models
• Configure provider routing and model fallbacks
• Build an agent application