Langfuse
App : A product you use directly
Open source LLM observability: traces, costs, evaluations, prompt management.
The go-to self-hostable solution, which remained MIT-licensed after the acquisition by ClickHouse.
Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.
What it does, concretely
When building an application on top of a language model, it is hard to know what is really happening inside: which calls are made, how much they cost, why a response is wrong. Langfuse addresses this by recording every interaction with the model, every data retrieval step or agent action, and making them inspectable in a centralised interface.
In practice, you integrate a Python or JavaScript SDK into your application, or use one of more than a hundred available integrations (LangChain, LlamaIndex, LiteLLM, Vercel AI SDK, OpenAI, and others). Langfuse then collects hierarchical traces capturing every LLM call, tool invocation, and retrieval step, along with associated cost and latency. You can then evaluate response quality via an LLM judge, code functions, user feedback, or manual annotation, and manage your prompts with versioning and one-click deployment.
Langfuse is aimed at engineering teams and developers building agents or LLM applications in production, whether working alone or in a team. The platform is used by more than 50,000 companies, including Fortune 50 members, and processes more than 90 billion observations per month.
However, Langfuse is not designed for non-technical or general consumer use: you need to be comfortable with code to get value from it. People who simply want to use a chatbot or an AI tool without building their own application have no use for it.
What it offers
- ·Hierarchical tracing of every LLM call, tool invocation, and retrieval step, with associated cost and latency.
- ·Prompt management with versioning, one-click deployment, rollback, and server- and client-side caching to avoid added latency.
- ·Response evaluation via LLM-as-a-judge, code functions, user feedback, or collaborative manual annotation.
- ·Datasets and benchmarks for testing the application before deployment and comparing experiments side by side.
- ·Built-in playground to test and iterate on prompts and model configurations directly from a problematic trace.
- ·Dashboards for monitoring cost, latency, and quality, with automated alerts.
- ·Full REST API with OpenAPI specification and typed SDKs for Python and JavaScript/TypeScript.
- ·Self-hosted deployment in minutes via Docker Compose, Kubernetes (Helm), or Terraform templates for AWS, Azure, and GCP.
In practice
- Price
- Freemium
- Commercial use
- Commercial use: yes
- Available on
- web, docker
- Licence
- MIT
Security and data
- Rating
- Security: nothing to flag
- Your data
- Local or cloud depending on use
- Known incidents
- Rachat par ClickHouse annoncé le 16 janvier 2026.
- Maintenance
- Active
- Last release
- 24 October 2026
Links and repository checked on 25 October 2026 · added on 1 May 2023
How to install Langfuse
git clone --depth=1 https://github.com/langfuse/langfuse.git && cd langfuse && docker compose up