Langfuse

Langfuse is an open-source LLM engineering platform that helps teams build, debug, and improve applications powered by large language models. It centers on tracing: capturing every LLM call and related step in an application, including retrieval, embeddings, and other API calls, so developers can see exactly what an agent or chain did during a run. Traces can be grouped into multi-turn sessions, visualized as agent graphs, and analyzed for cost, latency, and quality across users. Alongside tracing, Langfuse offers prompt management with version control and label-based deployment, so teams can update and roll back prompts without redeploying application code, plus a built-in LLM playground for interactive testing. Its evaluation tools support several methods — LLM-as-a-judge scoring, custom code evaluators, user feedback, and manual labeling — which can run in production or against curated datasets during development. Langfuse integrates with more than 100 frameworks and providers, including OpenAI, LangChain, LlamaIndex, and LiteLLM, and supports OpenTelemetry natively. Native SDKs exist for Python and JavaScript, and the full platform is available through a public API. It is fully open source and can be self-hosted, or run as a managed cloud service with a free tier for smaller projects. Exact enterprise pricing is not listed on the site.