Langfuse is an open-source AI engineering platform (MIT-licensed core, 24,000+ GitHub stars) that provides production observability, prompt management, evaluation, and experimentation for LLM applications. It captures traces across OpenAI, LangChain, LlamaIndex, and other frameworks, tracking latency, cost, and errors. The platform connects tracing to prompts, evals, annotation queues, and datasets in a single workflow. It can be self-hosted with no seat caps or usage limits, or used via the managed cloud with a generous free tier. Key features: - End-to-end LLM tracing: latency, cost, token counts, errors across all major frameworks - Prompt management with versioning, A/B testing, and environment-based rollouts - Evaluation framework: LLM-as-judge, custom evaluators, and human annotation queues - OpenTelemetry integration for standardized observability - Datasets and experiment tracking for systematic prompt improvement - Self-hostable with MIT license: no seat caps, no retention limits, no usage caps
Hobby: free (50k units/month, 30-day retention, 2 users). Core: $29/month (100k units, 90-day retention, unlimited users). Pro: $199/month (3-year retention, SOC2/ISO27001). Enterprise: $2,499/month (audit logs, SCIM, SLA). Overage: $8/100k units.