AI & LLM integrations

Trace the model calls, agent steps, and tool calls in your AI application. Instrument the SDK or library you already use, or connect a hosted telemetry source such as OpenRouter Broadcast. The resulting spans show up in Logfire with prompts, tokens, cost, and latency where the integration captures them.

Looking for an agent framework?

These pages cover model-provider SDKs, LLM libraries, and hosted model gateways. If you build on an agent framework (CrewAI, LangGraph, Agno, VoltAgent, and more, across Python, TypeScript, Go, Rust, and .NET), see Agent Frameworks for per-framework setup and which Logfire views each populates.

Observability integrations vs. the Logfire AI Gateway

These pages send telemetry to Logfire; they do not route model traffic through Logfire. To route calls through Logfire for spending caps and provider failover, see Connect an agent to the AI Gateway.

Pydantic AIOpenAIOROpenRouter BroadcastGoogle Gen AIAnthropicLangChainLiLiteLLMDSPyMCPLlamaIndexMirascopeMaMagentic

Not listed?

Any OpenTelemetry-compatible LLM SDK can send to Logfire. See Alternative clients for the generic pattern.