Langfuse

Langfuse is a purpose-built observability platform for AI engineering teams building LLM applications and agentic systems. It provides end-to-end tracing to capture every operation—LLM calls, retrieval steps, tool executions, and custom logic—with full timing and metadata context. This enables rapid debugging and optimization of production AI workflows.

Highlights

• Complete tracing of LLM requests with prompt, response, and token tracking
• Prompt management and versioning for collaborative iteration
• LLM-as-Judge evaluation and human feedback collection for quality assessment
• Cost and token usage monitoring across multiple models
• Open source (MIT) and self-hostable with ClickHouse backend

The platform natively understands LLM concepts like token usage, model parameters, and evaluation scores, making it framework and model agnostic. Teams use Langfuse to confidently ship and iterate on production AI systems with visibility and control. The free tier includes 50,000 units monthly, making it accessible for evaluation before scaling.