Langfuse

Langfuse is a platform for tracing, evaluating, and improving AI agents from prototype through production. It connects observability, prompts, evaluations, datasets, experiments, and human feedback in one workflow.

It captures hierarchical traces for LLM calls, tool invocations, and retrieval steps, with filtering by user, session, cost, latency, and custom metadata. Teams can run LLM-as-a-judge, heuristic, or human evaluations, manage prompts with deployments and rollbacks, compare models in a playground, and monitor cost and latency.

Langfuse works with languages and frameworks that support OpenTelemetry instrumentation, alongside native Python and TypeScript SDKs and more than 100 integrations. A free starting option is available, and the project is open source.

Highlights

• Tracing and monitoring for AI agents and LLM applications
• Evaluations using LLM judges, heuristics, or human review
• Prompt management with deployments and rollbacks
• Experiments, datasets, playground comparisons, and human annotation
• Integrations with agent frameworks, model providers, and developer tools