Pydantic Logfire
Logfire is the infrastructure your agents run on. It traces every agent run, helps prove fixes with evals, and runs the code agents write in Agent Sandboxes. It combines a shared evaluation workspace with production tracing and can be hosted by Pydantic or self-hosted on Enterprise.
Provider: Pydantic
- encryption_at_rest: Access tightly controlled
- free_tier_records: 10M telemetry records per month
- gdpr: Data processing addendum available
- hipaa: BAA available
- penetration_testing: Independently tested every year
- pydantic_validation_downloads_all_time: 18B+ on PyPI
- self_hosting: Available on Enterprise
- soc2_type_2: Independently audited
Buying
- Public pricing
- Unknown
- Enterprise offering
- Unknown
- Founded
- Unknown
Personal: $0/month with 10M records and 30-day retention. Team: $49/month with 10M records, then $2 per million additional records. Growth: $249/month with up to 90-day retention. There is no separate score fee; evaluation records count toward the same monthly record allowance. Model and judge inference costs are separate. Enterprise supports self-hosting.
Capabilities
- Trace every agent run
- Offline and online evals
- LLM-as-a-judge and custom evaluators
- Shared datasets and experiment comparison
- Agent Sandboxes for secure isolated code execution
- Model routing through Pydantic AI Gateway
- OpenTelemetry/OTLP ingest and export
- Python and TypeScript SDKs
- PostgreSQL-compatible SQL querying of evaluation records and traces
- Hosted by Pydantic or self-hosted on Enterprise
Categories
- Agent sandboxes
- LLM evaluation tools
- LLM observability