Pydantic Logfire

Instrumentation in Logfire is one import and a decorator away for most Python apps, which is the Pydantic team's pitch: an observability platform built entirely on OpenTelemetry, so traces, metrics, and logs flow from the MIT-licensed Python and JavaScript SDKs to any OTel-compatible backend, not only Pydantic's own. The hosted platform adds what generic APM tools lack for LLM work: conversation panels that render agent runs as chat transcripts, token and cost accounting per model, tool-call inspection, and plain SQL over trace data for ad hoc analysis. Auto-instrumentation covers FastAPI, HTTPX, OpenAI, Anthropic, and Pydantic AI among dozens of integrations, plus Python-specific extras like event-loop telemetry and database query profiling. The SDK installs from PyPI with pip install logfire; the server side is closed source, offered as SaaS with a free tier and paid plans, and self-hosting is available only under an enterprise license. For teams already building on Pydantic AI or FastAPI it is the lowest-friction route to production-grade traces.