Langfuse

The product is an open source platform for tracing, evaluating and improving AI agents and LLM applications, using production data to debug behavior, analyze cost and latency, and manage prompts and evaluations. It sells to developers, engineering teams and data teams in B2B companies from startups to enterprises building production-grade LLM applications. It is positioned as an open source alternative to legacy observability platforms, differentiated by MIT licensing, self-hosting and data control, and is delivered as SaaS and self-hosted software via Docker, Kubernetes and Terraform on AWS, GCP and Azure.

Key features

• LLM tracing and observability
• AI agent evaluation
• Prompt management
• Evaluation metrics and scoring
• Regression datasets
• Token cost and latency breakdown
• Async ingestion via Redis queue
• ClickHouse OLAP storage
• S3 blob storage export
• REST APIs and Query SDK
• In-app assistant for investigations
• SKILL.md for coding agents
• CLI for terminal workflows
• MCP server for IDE agents
• Edge-cached prompts