About Logfire
Pydantic Logfire is an observability platform built on OpenTelemetry and aimed at AI applications. It records the model call, agent steps, API requests and database queries in a single trace, so a failure can be followed across the whole application rather than only the LLM layer. Logs, spans and metrics can be queried with Postgres flavored SQL in the interface or through MCP. Evaluations run inside the trace, with scores attached to the span they measure and runs compared against each other.
It is built for developers and teams shipping LLM apps, agents and RAG systems, though it also covers ordinary services and hosts. SDKs are provided for Python, JavaScript and TypeScript, and Rust, and any OpenTelemetry compatible language works. The vendor lists SOC 2 Type II, HIPAA support, EU data residency, PII scrubbing in the SDK and a self hosted option. An AI gateway supports bringing your own provider credentials.
Pricing is based on records. The Personal plan is free with 10 million records per month, three projects, 30 day retention and one seat. Team is $49 per month with 10 million records included, additional records at $2 per million, and five seats with extra seats at $25. Growth is $249 per month. Dedicated, Enterprise and self hosted options are priced on request.