Langfuse
Open-source tracing, evals, and prompt management for LLM apps.
Visit Langfuse Free Hobby cloud tier and self-hosting; Core from $59/mo
What is Langfuse?
Langfuse is an open-source observability platform for LLM applications: trace every call, track cost and latency, run evaluations, manage prompt versions, and collect user feedback. Self-host for free or use the managed cloud.
Key features
- Traces for multi-step agent and RAG pipelines
- Cost and token tracking per user, feature, or session
- Prompt management with versioning and A/B testing
- LLM-as-judge and human evaluation workflows
- Integrations with LangChain, LlamaIndex, OpenAI, and Anthropic SDKs
Pros
- Fully open source with a generous free tier
- Framework-agnostic — works with plain SDK calls too
- Prompt management and evals in the same tool as tracing
Cons
- Self-hosting requires Postgres and ClickHouse to run well
- UI is dense for non-technical stakeholders
- Younger than LangSmith in some enterprise features
Best for
Teams that want observability without vendor lock-inDevelopers not using LangChain who still need tracingStartups controlling LLM spend
Related comparisons
Alternatives to Langfuse
LangSmith
LLM observability + evals from the LangChain team — production tracing for AI apps.
FreemiumFree 5k traces/mo, Plus $39/mo
Released July 2023
Helicone
Open-source LLM observability — one-line proxy, full request logs, cost tracking.
FreemiumFree 100k requests/mo, Pro $25/mo
Released February 2023
Vellum
LLM dev platform — prompt versioning, evals, monitoring in one tool.
PaidFrom $500/mo
Released July 2023