Connect frameworks to the AI Gateway

Connect the framework, prompts, tools, and agent workflow you already use to the Logfire AI Gateway.

Gateway vs. instrument

These guides route your model calls through the gateway for spending caps, failover, and shared keys. If you only want to see and debug the calls you already make, you don’t need the gateway: instrument a framework instead.

Each guide shows the two client settings you need to change: the API key and gateway URL. You can point them at one model provider or a gateway endpoint, which can fail over to another provider or distribute requests across several providers.

Before you start

1. Follow Send your first Gateway request to enable Gateway, choose a provider or endpoint, and get a project-scoped Gateway API key. An organization admin must enable Gateway and configure a provider.

2. Set the Gateway API key in the terminal where you will run the example. Keep it out of source control:

   Terminal

   export LOGFIRE_GATEWAY_API_KEY="..."
   

3. On the Gateway Connect tab, select a route and model. Copy the gateway URL and model name into the example.

The examples use the OpenAI-compatible route for the US region, https://gateway-us.pydantic.dev/proxy/openai. For the EU region, use gateway-eu instead. For a self-hosted organization, copy the URL from the Connect tab.

Most examples use an OpenAI-compatible client. The provider or gateway endpoint you select must support the OpenAI request format and the model name in the example. Provider-native APIs, such as Anthropic Messages, require that provider’s client. See Connect an SDK for both patterns.

Model data passes through Logfire

These settings send prompts, tool inputs, and model responses through the Logfire AI Gateway and the selected model provider. If gateway telemetry is enabled, Logfire records the model, latency, token usage, and conversation content in your selected project. Calls to built-in providers count toward your gateway spend.

Choose your framework

Framework or SDK	Languages
Agno	Python
Genkit	Go, TypeScript
Google ADK	Go, Python
Haystack	Python
Instructor	Python
LangChain	TypeScript, Python
LlamaIndex	Python
Mastra	TypeScript
Microsoft Agent Framework	.NET, Python
OpenAI Agents SDK	TypeScript, Python
OpenAI SDK	.NET, Go, TypeScript, Python
Semantic Kernel	.NET, Python
Smolagents	Python
Strands Agents	TypeScript, Python
Vercel AI SDK	TypeScript
VoltAgent	TypeScript

Next steps

Read the AI Gateway overview to learn how providers, gateway endpoints, spending limits, and telemetry work together.