🪢 Open source LLM engineering platform. Observability, metrics, evals, prompt management, testing, prompt playground, datasets, LLM evaluations -- 🍊YC W23 🤖 integrate via Typescript, Python / Decor…



Langfuse GitHub Banner

Langfuse uses Github Discussions for Support and Feature Requests.
We're hiring. Join us in Backend Engineering, Product Engineering, and Developer Relations.

MIT License Y Combinator W23 Docker Image langfuse npm package langfuse Python package on PyPi


Unmute video for voice-over




  • Experiments: Track and test app behaviour before deploying a new version

Get started

Langfuse Cloud

Managed deployment by the Langfuse team, generous free-tier (hobby plan), no credit card required.

» Langfuse Cloud

Localhost (docker)

# Clone repository
git clone
cd langfuse

# Run server and database
docker compose up -d

→ Learn more about deploying locally

Self-host (docker)

Langfuse is simple to self-host and keep updated. It currently requires only a single docker container. → Self Hosting Instructions

Templated deployments: Railway, GCP Cloud Run, AWS Fargate, Kubernetes and others

Get Started

API Keys

You need a Langfuse public and secret key to get started. Sign up here and find them in your project settings.

Ingesting Data · Instrumenting Your Application

Note: We recommend using our fully async, typed SDKs that allow you to instrument any LLM application with any underlying model. They are available in Python (Decorators) & JS/TS. The SDKs will always be the most fully featured and stable way to ingest data into Langfuse.

You may want to use another integration to get started quickly or implement a use case that we do not yet support. However, we recommend to migrate to the Langfuse SDKs over time to ensure performance and stability.

See the → Quickstart to integrate Langfuse.


SDKPython, JS/TSManual instrumentation using the SDKs for full flexibility.
OpenAIPython, JS/TSAutomated instrumentation using drop-in replacement of OpenAI SDK.
LangchainPython, JS/TSAutomated instrumentation by passing callback handler to Langchain application.
LlamaIndexPythonAutomated instrumentation via LlamaIndex callback system.
HaystackPythonAutomated instrumentation via Haystack content tracing system.
LiteLLMPython, JS/TS (proxy only)Use any LLM as a drop in replacement for GPT. Use Azure, OpenAI, Cohere, Anthropic, Ollama, VLLM, Sagemaker, HuggingFace, Replicate (100+ LLMs).
APIDirectly call the public API. OpenAPI spec available.

Packages that integrate with Langfuse:

InstructorLibrary to get structured LLM outputs (JSON, Pydantic)
MirascopePython toolkit for building LLM applications.
AI SDK by VercelTypescript SDK that makes streaming LLM outputs super easy.
FlowiseJS/TS no-code builder for customized LLM flows.
LangflowPython-based UI for LangChain, designed with react-flow to provide an effortless way to experiment and prototype flows.

Questions and feedback

Ideas and roadmap

Support and feedback

In order of preference the best way to communicate with us:

Contributing to Langfuse

  • Vote on Ideas
  • Raise and comment on Issues
  • Open a PR - see for details on how to setup a development environment.


This repository is MIT licensed, except for the ee folders. See LICENSE and docs for more details.


GET API to export your data

GET routes to use data in downstream applications (e.g. embedded analytics). You can also access them conveniently via the SDKs (docs).

Security & Privacy

We take data security and privacy seriously. Please refer to our Security and Privacy page for more information.


By default, Langfuse automatically reports basic usage statistics of self-hosted instances to a centralized server (PostHog).

This helps us to:

  1. Understand how Langfuse is used and improve the most relevant features.
  2. Track overall usage for internal and external (e.g. fundraising) reporting.

None of the data is shared with third parties and does not include any sensitive information. We want to be super transparent about this and you can find the exact data we collect here.

You can opt-out by setting TELEMETRY_ENABLED=false.