Langfuse
Open-source LLM engineering platform for tracing, evals, and prompt workflows.
Open source
Best when teams want self-hostable observability with datasets, scores, and prompt management in one stack.
Selection advice
Choose Langfuse when you need MIT-licensed tracing plus eval workflows without locking into a single framework vendor.
Best for
- self-hosted agent tracing
- production eval loops
- prompt versioning with traces
Not ideal for
- teams that only need a hosted LangChain-native workflow
- projects with no appetite to operate observability infrastructure
Core concepts
tracesobservationsscoresdatasetsprompts
Minimal implementation shape
Instrument one agent run, inspect spans for tool calls and retrieval, tag failures as scores, and export recurring cases into a dataset.