Langfuse: A self-hostable LLM observability and evaluation platform built on ClickHouse
GitHub describes it as 🪢 Open source AI engineering platform: LLM evals, observability, metrics, prompt management, playground, datasets. Integrates with OpenTelemetry, LangChain, OpenAI SDK, LiteLLM, and more. 🍊YC W23. The repository metadata lists TypeScript as its primary language. The metadata lists the NOASSERTION license. This article stays within the project description and details documented in the GitHub repository README.
At a glance
- What is it?
- Langfuse is an open source platform for tracing, evaluating, and managing prompts in LLM applications. It can be self-hosted with Docker Compose or Helm, and it integrates with OpenAI, LangChain, and OpenTelemetry.
- Who is it for?
- Adopt Langfuse if you need a self-hosted LLM observability and evaluation platform and you are comfortable running or maintaining a ClickHouse-backed stack. It is a poor fit for teams that want a fully managed service without operational overhead.
- Can I use it commercially?
- Check first. The repository uses a licence we do not classify automatically, so read its LICENSE file before any commercial use.
- Is it still maintained?
- Yes. The repository received new commits within the last day.
- What is it written in?
- Mainly TypeScript, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- langfuse/langfuse37,453 ★ · +4,399
| Repository | 2026-08-13 | 2026-11-15 |
|---|---|---|
| langfuse/langfuse | 33.1k | 37.5k |
4,399 Stars in the last 94 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
What Langfuse actually does
Langfuse is an open source platform for LLM engineering. It covers observability, evaluations, prompt management, datasets, and a playground. The core problem it solves is the lack of a unified view of LLM application behavior. Without it, teams juggle raw logs, ad-hoc scripts, and manual prompt versions. Langfuse gives you a single place to trace calls, label bad outputs, run evals, and iterate on prompts. It is aimed at developers and teams building AI applications, from small startups to larger organizations that want to self-host. The README says it can be self-hosted in minutes and is battle-tested.
How the platform is architected
The architecture is visible from the README and the repository layout. Langfuse is built with TypeScript as the primary language. Its default compose file runs ClickHouse alongside Postgres, Redis and MinIO, and ClickHouse is column-oriented and designed for high-volume analytics. That choice suggests Langfuse is optimized for ingesting large numbers of trace events and running analytical queries over them. The platform includes a web UI for inspecting traces and sessions, an API with OpenAPI spec, and typed SDKs for Python and JS/TS. The data flow is straightforward: your application sends trace data via an SDK or an integration, Langfuse stores it, and you view or query it through the UI or API. The README also mentions strong caching on server and client side for prompt management, which implies that prompt retrieval is designed to avoid adding latency to your application.
Getting it running: Docker Compose and beyond
The README gives a concrete path for local deployment. You clone the repository with git clone --depth=1 https://github.com/langfuse/langfuse.git, then cd langfuse, then run docker compose up. That gets a local instance running in about five minutes, according to the README. For production, the preferred method is Kubernetes via Helm. There are also Terraform templates for AWS, Azure, and GCP. The README points to the self-hosting documentation for architecture and configuration options. The Docker Compose path is the quickest way to evaluate the platform, but it is not a production setup.
Integrations: where the value multiplies
Langfuse ships with several integration layers. The SDKs for Python and JS/TS give you manual instrumentation. There is an OpenAI SDK integration, and LangChain is also listed as an integration. OpenTelemetry is also listed as an integration, which is important because it allows you to unify LLM traces with other telemetry in your stack. The README mentions LiteLLM as well. These integrations matter because they reduce the friction of adoption. If you already use OpenAI or LangChain, you can get traces flowing with minimal code changes. The trade-off is that each integration may have its own version compatibility, and you need to verify that the integration supports the exact features you need.
Evaluations and datasets: the workflow glue
Evaluations are a core feature. Langfuse supports LLM-as-a-judge, code evaluators, user feedback collection, manual labeling, and custom evaluation pipelines via APIs and SDKs. That is a broad range, but it also means you must decide which method fits your workflow. Datasets let you build test sets and benchmarks for pre-deployment testing and continuous improvement. The README says datasets integrate with LangChain and LlamaIndex. This is where Langfuse goes beyond simple observability. It provides a structured way to run experiments and compare results. The playground lets you iterate on prompts and model configurations directly from a trace. The connection between tracing and evaluation is the main strength: you can spot a bad result, jump to the playground, tweak the prompt, and then run an eval on a dataset.
Licensing and maintenance considerations
The repository license field is NOASSERTION, meaning GitHub did not report a standard license identifier, while the root package.json declares the license as MIT. The README links to a LICENSE file, and the repository also has an ee/ directory, so read the LICENSE file in the repo before committing. The project is actively maintained, with recent releases v4.72.0 (2026-10-29), v4.71.0 (2026-10-27) and v3.225.15 (2026-10-27), and a last push on 2026-11-01, so development is active. The README also notes that since January 2026, Langfuse is part of ClickHouse. That could mean the project has corporate backing, but it does not guarantee license clarity. Self-hosting also means you are responsible for upgrades, backups, and ClickHouse maintenance. The self-hosting update list is offered for release notifications, which suggests that upgrades are a regular event.
Limitations and when to look elsewhere
One limitation is the operational burden. Running ClickHouse is not trivial. It is a distributed database with its own tuning and scaling considerations. For small teams, that might be overkill. The README says self-hosting is possible in minutes, but that is for local Docker Compose, not for a production deployment. Another limitation is license clarity: GitHub reports NOASSERTION while package.json declares MIT, so if your organization has strict open source compliance, read the LICENSE file first. As an alternative, consider OpenTelemetry with a backend like Jaeger or Grafana Tempo for tracing. That approach is more generic and does not provide built-in evals or prompt management. You would have to build those yourself. Langfuse is the right tool when you want an integrated LLM-specific platform, not a general-purpose tracing backend.
Editorial conclusion
Adopt Langfuse if you need a self-hosted LLM observability and evaluation platform and you are comfortable running or maintaining a ClickHouse-backed stack. It is a poor fit for teams that want a fully managed service without operational overhead. Before adopting, verify the exact license terms in the repository, confirm that the self-hosted version includes all features you need (some may be cloud-only), and run a pilot with your own OpenAI or LangChain workload to check trace volume and query performance. If you need a lighter-weight tracing tool, consider OpenTelemetry with a backend like Jaeger, but expect to build evaluation and prompt management yourself.
Frequently asked questions
Who owns Langfuse?
The project says that since January 2026 it is part of ClickHouse, and it says the platform is made with the ClickHouse open source database. ClickHouse also appears as one of the four services the default compose file waits on before starting the worker.
Is Langfuse free or paid?
There are two deployment paths. Langfuse Cloud is a managed deployment by the Langfuse team with a generous free tier and no credit card required, and self-hosting runs the same platform on your own infrastructure through Docker Compose, Helm, or Terraform templates for AWS, Azure and GCP. The root package.json declares the license as MIT.
how to install langfuse locally
The local path is a Docker Compose run on your own machine, which the README puts at 5 minutes: clone the repository, change into the langfuse directory, and run docker compose up. Before the stack is reachable, replace the credential placeholders marked CHANGEME in the compose file, which include DATABASE_URL, SALT and ENCRYPTION_KEY.
how to use langfuse for prompt management
Prompt Management lets you centrally manage, version control and collaboratively iterate on your prompts, and the README attributes the lack of added latency to strong caching on both the server and the client side. The LLM Playground is the companion surface for iterating on prompts and model configurations.
how to use langfuse api
The platform is used to power bespoke LLMOps workflows through its building blocks exposed via the API. An OpenAPI spec, a Postman collection, and typed SDKs for Python and JS/TS are available.
Official sources
Where this project is recommended
Where developers are discussing it
Posts on Hacker News, dev.to and Lobsters that link to this repository, found by our scan of those communities. 1 in the last 30 days; newest first.
- Show HN: Langfuse AI Gateway, open source rust gateway to trace every LLM callHacker News · Nov 13, 2026 · 9 points · 1 comments
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/langfuse-langfuse)Star history chart for your README
[](https://hysenlabs.com/projects/langfuse-langfuse)Was this useful?