Langfuse is the leading open-source LLM observability platform. It provides tracing, prompt management, evals, datasets, and usage analytics — all with a self-hostable option for teams that need data residency or air-gapped deployments. Framework-agnostic: works with LangChain, LlamaIndex, Instructor, or raw API calls.
Add the Langfuse SDK to your app and wrap LLM calls with a decorator or use their auto-instrumentation. Every trace is captured with structured metadata. The dashboard provides filtering, aggregation, and drill-down. Prompt management lets you version prompts and deploy them without redeploying code. Evals can run LLM-as-judge or custom Python functions against datasets.
Self-host: free, MIT licensed. Cloud: Free tier (50k observations/month), Pro $59/month, Team custom. Dedicated hosting available for enterprise.
Teams needing self-hosting, compliance-sensitive deployments, and framework-agnostic observability. Popular in European and regulated industries.