Langfuse

Langfuse builds an open source AI engineering platform that helps teams trace, evaluate, and improve their AI agents in production. It integrates observability, prompt management, evaluations, and analytics to support the full lifecycle of AI applications. The product is designed for developers who want control over data and open standards, with self-hosting and community involvement emphasized.

Langfuse provides a single platform that combines tracing, evaluation, prompts, and experiments to understand behavior, test improvements, and ship higher quality agents. It supports production data usage and collaboration on fixes, aiming to reduce cost and latency while increasing confidence in deployments.

Langfuse advocates open standards and data portability, highlighting MIT license and OSS community engagement. It offers APIs and exports, and notes that self-hosting is possible with Docker, Kubernetes, or cloud configurations.

Langfuse presents an integration-focused approach, with features spanning observability, prompt management, evaluations, and playground-style testing. It emphasizes that all product features are MIT licensed and that the platform can scale to large volumes of observations and company deployments. The home page also notes a focus on developers and agents, with Halo and MCP tooling mentioned as part of the ecosystem.

Use cases

• Trace every LLM call with cost and latency.
• Manage prompts and versions across environments.
• Run production evaluations and online/offline tests.