langfuse/Langfuse
🪢 Open source LLM engineering platform: LLM Observability, metrics, evals, prompt management, playground, datasets. Integrates with OpenTelemetry, Langchain, OpenAI SDK, LiteLLM, and more. 🍊YC W23
Langfuse is an open-source LLM engineering platform that gives developers tracing, evals, prompt management, and analytics for AI apps. It plugs into OpenAI SDK, LangChain, LlamaIndex, and LiteLLM, so you can see model calls, debug failures, and measure quality from one dashboard.
Our Review
Langfuse came out of Y Combinator (W23) and has grown into one of the most-used open-source stacks for running LLM apps in production, with 28,000 GitHub stars. The team open-sourced the core platform in 2025, so teams can self-host the same tracing and evaluation tooling that powers Langfuse Cloud.
What Langfuse does:
- •Tracing capture every LLM call, tool call, and chain step as a nested trace, so you can see exactly what your app did and why.
- •Evals score outputs with model-based, custom, or human evaluations and track quality over time on datasets.
- •Prompt management version, deploy, and roll back prompts from a central place without redeploying your app.
- •Playground test prompts and models interactively, then save the good ones straight to prompt management.
- •Analytics and dashboards track cost, latency, and quality metrics across models, users, and releases.
- •Broad integrations works with the OpenAI SDK, LangChain, LlamaIndex, LiteLLM, and OpenTelemetry.
Getting started:
Use Langfuse Cloud for a hosted start, or self-host with Docker following the deployment docs. Add the SDK (Python or JS), wrap your LLM calls, and traces appear in the dashboard. Docs at langfuse.com/docs.
Limitations:
Langfuse uses a mixed license: the core platform is MIT, but some enterprise features sit under a separate commercial license, so a fully self-hosted deployment may not include everything Langfuse Cloud offers. Self-hosting means running and scaling the stack (it uses Postgres and ClickHouse) yourself. The breadth of features across tracing, evals, prompts, and analytics is more than a small project may need.
Our Verdict
Langfuse is the default open-source choice for LLM observability in 2026. If you are shipping an AI feature and flying blind on what your model actually does, why it fails, and what it costs, Langfuse gives you traces, evals, and analytics in one place, with 28,000 stars and a YC pedigree behind it.
For developers, integration is the strong point: wrap your existing OpenAI, LangChain, LlamaIndex, or LiteLLM calls and traces show up automatically, no rearchitecting required. Prompt management and the playground let you iterate on prompts without code deploys, and the eval tooling turns a vague sense of quality into measured scores on real datasets.
Skip Langfuse if you want zero infrastructure and your needs are simple; a hosted analytics tool or basic logging may be enough, and self-hosting Postgres plus ClickHouse is real operational work. If you only need request logging without evals or prompt management, a lighter tool is less to run.
Frequently Asked Questions
What is Langfuse?
Langfuse is an open-source LLM engineering platform, started by a Y Combinator (W23) team, for observing and improving AI applications. It provides tracing of model and tool calls, evaluations, prompt management, a testing playground, and cost and quality analytics, all in one dashboard, and it integrates with the OpenAI SDK, LangChain, LlamaIndex, and LiteLLM.
Is Langfuse free and open source?
Langfuse is open source as of 2026, with its core platform under the MIT license, and you can self-host it for free. Some enterprise features are under a separate commercial license, and Langfuse also offers a paid hosted Cloud plan. For most teams, the free self-hosted core covers tracing, evals, and prompt management.
How does Langfuse integrate with my LLM app?
Langfuse provides Python and JavaScript SDKs plus native integrations for the OpenAI SDK, LangChain, LlamaIndex, LiteLLM, and OpenTelemetry. In most cases you wrap or decorate your existing model calls, and traces, costs, and latencies appear in the dashboard without rearchitecting your app. You can also send data through its standard OpenTelemetry endpoint.
What is the difference between Langfuse and LangSmith?
Both are LLM observability and evaluation platforms. Langfuse is open source with a self-hostable MIT core, while LangSmith is a closed-source commercial product tied to the LangChain ecosystem. Choose Langfuse when you want self-hosting, open source, and framework-agnostic integrations; choose LangSmith when you are all-in on LangChain and prefer a fully managed service.
Can I self-host Langfuse?
Yes. Langfuse is designed to be self-hosted as of 2026, and the team documents Docker-based deployment. It runs on Postgres and ClickHouse for storage, so you provision and scale those yourself. Self-hosting gives you full data control at the cost of running the infrastructure; Langfuse Cloud is the managed alternative if you would rather not.
How do I install Langfuse?
Visit the GitHub repository at https://github.com/langfuse/langfuse for installation instructions.
What license does Langfuse use?
Langfuse uses the MIT license.
What are alternatives to Langfuse?
Explore related tools and alternatives on My AI Guide.
Open source & community-verified
MIT licensed: free to use in any project, no strings attached. 28,459 developers have starred this, meaning the community has reviewed and trusted it.
Reviewed by My AI Guide for relevance, quality, and active maintenance before listing.
Topics