Langfuse
Langfuse is an open-source LLM engineering platform for developing, monitoring, evaluating and debugging AI applications. It traces model calls, retrieval and agent actions, manages prompts with versions, runs evaluations from LLM-as-a-judge to human labelling, and keeps datasets and a playground. It is MIT outside its enterprise folders, self-hosts with Docker or Kubernetes, has Python and TypeScript SDKs, and has been part of ClickHouse since January 2026.
What does Langfuse do?
Langfuse is "an open source LLM engineering platform" that "helps teams collaboratively develop, monitor, evaluate, and debug AI applications." Tracing captures "LLM calls and other relevant logic in your app such as retrieval, embedding, or agent actions", and when you see a bad result you can jump from the trace to the playground to iterate on it. Prompt management keeps prompts under version control, cached on server and client so it adds no latency.
Evaluations cover "LLM-as-a-judge, Code evaluators, user feedback collection, manual labeling, and custom evaluation pipelines", and datasets hold test sets and benchmarks. It has an OpenAPI spec, a Postman collection and typed SDKs for Python and JavaScript and TypeScript. Its integrations table lists OpenAI, LangChain, LlamaIndex, Haystack, LiteLLM, the Vercel AI SDK and Mastra, and a second table lists 19 packages that integrate with it, including Ollama, Amazon Bedrock, DSPy, CrewAI, Dify and Open WebUI. The README says "since January 2026 we're part of ClickHouse."
Key facts
- Licence: MIT except the ee, web/src/ee and worker/src/ee folders, which are under the Langfuse Enterprise license; the ee licence states the MIT core runs without it.
- Copyright ClickHouse, Inc.; the README says Langfuse joined ClickHouse in January 2026.
- Tracing of model calls, retrieval, embeddings and agent actions, linked to a playground.
- Prompt management with version control, cached server and client side.
- Evaluations: LLM-as-a-judge, code evaluators, user feedback, manual labelling and custom pipelines; datasets for tests and benchmarks.
- APIs and SDKs: OpenAPI spec, Postman collection, typed Python and JS/TS SDKs; integrations include OpenAI, LangChain, LlamaIndex, LiteLLM, the Vercel AI SDK, Ollama, Bedrock and CrewAI.
- Install: Docker Compose on your own machine in about 5 minutes by the README, a single VM, Kubernetes with Helm for production, or Terraform for AWS, Azure and GCP. Langfuse Cloud has a free tier with EU and US regions.
- Latest tagged release when read: v4.38.0 on 2026-09-17; 34,745 stars on 2026-09-17.
What does it replace, and where does it fall short?
Langfuse replaces LangSmith for a team that wants tracing, prompt management and evaluations for its AI application on its own infrastructure, independent of any one framework, with the core under MIT. It is the most widely used entry on the open-source alternatives to LangSmith page, by stars.
Where it falls short: some features sit in the ee folders under a paid licence, so check which you need. It is now owned by ClickHouse, a commercial company, which the README states. For a single Apache-2.0 licence across the whole repository, Opik.
How does Langfuse run?
Run it with Docker Compose on your own machine, on a single VM, on Kubernetes with the Helm chart, which the README calls the preferred production deployment, or with Terraform on AWS, Azure or GCP. Langfuse Cloud is the managed version with a free tier, no card required, in EU and US regions.
Who is Langfuse for?
A team shipping an LLM application that needs traces, versioned prompts and evaluations, self-hosted, across whichever framework it uses. A team that wants no enterprise folders at all should look at Opik.
What limits does the README state?
None stated in the README.
Questions people ask
Is Langfuse open source?
Its core is, under MIT: the README says the repository is MIT licensed except for the ee folders, and the enterprise licence file says the MIT core can be used and run without it. Features in the ee folders need a Langfuse Enterprise licence.
Who owns Langfuse?
The README says Langfuse has been part of ClickHouse since January 2026, and the licence copyright line reads ClickHouse, Inc. The core licence remains MIT, and the enterprise folders keep their own licence, so the split described above is unchanged by the acquisition.
Can I self-host Langfuse?
Yes. The README describes Docker Compose on your own machine in about five minutes, a single VM, Kubernetes with Helm as the preferred production setup, and Terraform templates for AWS, Azure and GCP. Langfuse Cloud is the managed alternative.
How does Langfuse compare with LangSmith?
Both trace, evaluate and manage prompts for LLM applications. Langfuse self-hosts with an MIT core and works across frameworks, with integrations for LangChain, LlamaIndex, LiteLLM, the Vercel AI SDK and others. The open-source alternatives to LangSmith page compares it with Phoenix and Opik.
Does Langfuse work with local models?
Yes, through its integrations: Ollama is listed as a local model integration, and its LiteLLM integration reaches Ollama, vLLM and other providers. Langfuse records what your application sends to the model, wherever the model runs.
Sources
- Langfuse README, LICENSE and ee/LICENSE: github.com/langfuse/langfuse, read 2026-09-17.
Compared with the others
On the open-source alternatives to LangSmith page, Langfuse is the pick for the most widely used, mit core. Tracing of model calls, retrieval and agent actions, prompt management with versions, evaluations and datasets, self-hosted with Docker, Kubernetes or Terraform.
Also on that page: Opik for one apache-2.0 licence, plus ci tests, Arize Phoenix for opentelemetry across four languages.
More llm observability and evaluation
Comet's Apache-2.0 LLM observability and evaluation platform, one licence across the whole repository, for agent tracing, LLM-as-a-judge evaluations, prompt management, production monitoring, CI tests and an MCP server.
Arize's AI observability platform built on OpenTelemetry for tracing, evaluations, datasets, experiments and prompt management, under the Elastic License 2.0, which forbids offering it as a hosted service.
Added September 17, 2026. Every claim above comes from the project's README, LICENSE or model card, read on September 17, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/langfuse/langfuse.
Running one of these inside your own environment, with your own data and your own security rules, is the kind of work Reveneau does. Read how a forward deployed engagement works.