Pydantic AI Gateway
Enterprise-ready AI model routing (that developers love)
One key for all your models with real-time monitoring and budget control that works.
Free to get started
Intelligent AI Workflows without the middleware bloat
Start with one line of code
from pydantic_ai import Agent agent = Agent( 'gateway/openai:gpt-5', instructions='Be concise, reply with one sentence.' ) result = agent.run_sync('Hello World') print(result.output) You can access multiple models with the same API key.
To use different models, change the model string gateway/<api_format>:<model_name> to other models offered by the supported providers.
Available Providers
| Provider | Available? | Built-in? | BYOK? Bring Your Own Key | Model families provided |
|---|---|---|---|---|
 Anthropic | Yes | Yes | Yes | Claude |
 AWS Bedrock | Yes | Yes | Yes | Nova, Claude, etc... |
 Google Vertex | Yes | Yes | Yes | Gemini, Claude |
 Groq | Yes | Yes | Yes | Open Source models |
 OpenAI | Yes | Yes | Yes | GPT |
 Chat & Responses | Yes | No | Yes | Any compatible provider |
 Azure | Soon... | No | Yes | GPT |
Provider
Model families
Anthropic
AWS Bedrock
Google Vertex
Groq
OpenAI
Chat & Responses
AzureCost control for AI Models, Native APIs underneath
What is the Pydantic AI Gateway?
Pydantic AI Gateway gives you cost control and intelligent routing without the abstraction overhead. Unlike traditional AI gateways that wrap providers in a universal schema, PAIG passes requests through in their native format. When providers like OpenAI, Anthropic, or Google ship new features, you can use them immediately.
The core is open source (AGPL-3.0) with a cloud dashboard through Pydantic Logfire, or self-host for enterprise. Start with BYOK or use our built-in providers for single-key access to all models. PAIG comes with smart routing defaults you can fully customize. Use with Pydantic AI for one-line integration or point any existing provider SDK at the gateway, your code stays the same.
Why use Pydantic AI Gateway?
One key, multiple models
Connect to models on OpenAI, Anthropic, Google, Groq, and AWS Bedrock with one API key. Access multiple providers instantly without swapping credentials.
Flexible spend control
Set limits on your terms. Choose your level of control: project, user, or API key. Pick your timeframe: daily, weekly, monthly, or total.
No schema translation
Requests flow in each provider's native format. New model features are accessible as soon as they are released by providers.
Send OpenTelemetry traces for full visibility
Log every request through Pydantic Logfire or any OTel backend. Create a centralized audit trail where no API usage can hide.
Self-hosting available immediately
Deploy to your Cloudflare account, or run on-premises with our consulting support. control the data and the infrastructure.
Enterprise without the wait
SSO with OIDC, granular permissions, and flexible deployment—ready now (not a roadmap promise).
Free to get started
Gateway pricing
Bring your own provider credentials (BYOK): free on every plan, no markup
Built-in providers: 5% markup on Personal and Team, 3% on Growth