Pydantic Logfire AI Gateway

Pydantic Logfire AI Gateway is an analytics tool. Pydantic AI Gateway routes requests to OpenAI, Anthropic, Google, Groq, Bedrock and Azure through one API key, keeping native provider formats while adding spend limits and OpenTelemetry tracing via Logfire.

Pydantic AI Gateway is a model routing layer from the team behind Pydantic and Pydantic Logfire. It gives developers a single API key for models from OpenAI, Anthropic, Google Vertex, Groq, AWS Bedrock and Azure, with spend limits and request monitoring built in. Teams that already hold provider credentials can bring their own keys instead of using the built-in providers. The main design choice is that requests pass through in each provider's native format rather than being translated into a universal schema. When a provider ships a new feature, it can be used straight away, because the gateway does not need to add support first. A generic Chat and Responses option also covers any compatible provider that is not on the built-in list. In practice, a developer using the Pydantic AI framework changes a model string to the form gateway/api_format:model_name, and switching models means editing that string. Code written against an existing provider SDK can instead be pointed at the gateway, leaving the rest of the application unchanged. Cost control is the other core function. Limits can be set per project, per user or per API key, over daily, weekly, monthly or total periods. Every request can be logged through OpenTelemetry traces to Pydantic Logfire or any other OTel backend, producing a central audit trail of API usage. The dashboard runs in the cloud through Logfire, and enterprises can self-host on their own Cloudflare account or run on-premises with consulting support. SSO with OIDC and granular permissions are listed as available now. It suits engineering teams and platform groups that want governance over multi-provider LLM usage without adopting a heavy abstraction layer. Compared with gateways that normalize every provider into one schema, it trades uniformity for immediate access to provider-specific features. Teams already using Pydantic AI or Logfire will find the integration the most direct, while others rely on the SDK redirect approach.