Langfuse

Langfuse is an open-source observability and evaluation platform for LLM applications, designed to help teams systematically debug, monitor, and improve their AI workflows. It provides end-to-end tracing, prompt management, and performance analytics so developers can understand how their application interacts with language models at every step. By centralizing this information, Langfuse makes it easier to iterate on prompts, models, and system design with data-driven insights.

Key features include detailed tracing of LLM calls, chains, and tools, allowing users to inspect inputs, outputs, latencies, and costs across complex pipelines. Langfuse supports evaluations (evals) with custom metrics, test runs, and regression checks to quantify quality changes over time. Its prompt management capabilities enable versioning, comparison, and reuse of prompts, while metrics dashboards surface trends in latency, error rates, token usage, and user-defined KPIs. The platform integrates with popular frameworks and providers such as LangChain, OpenAI, LlamaIndex, LiteLLM, and others, making it straightforward to add observability without major code changes.