LangFuse.
Open-source LLM engineering platform — Berlin-based, GDPR-clean by default.
What it is.
LangFuse is the open-source LLM engineering platform built in Berlin. MIT licensed, Y Combinator W23 alum, and the default observability tool for European buyers who need EU data residency without contract negotiation. The product covers traces, prompts, evals, and datasets in one open-source codebase, with a managed cloud and enterprise tier above it.
Where it fits.
EU-headquartered teams and regulated buyers anywhere that need GDPR-clean LLM observability they can self-host. Also the default choice for teams that want the LangSmith feature set under an MIT licence. Strong for indie developers, EU enterprises, and any team where data residency is a procurement gate.
- MIT open-source — fully self-hostable
- EU data residency by default on the cloud tier
- Feature parity with LangSmith on traces and prompts
- Self-hosted deployment requires Postgres and infra ownership
- Eval tooling lighter than Braintrust's purpose-built workflow
Frequently asked.
Is LangFuse a true open-source alternative to LangSmith?
Yes. MIT licence covers the full feature set. LangSmith is closed-source with a self-hosted enterprise option; LangFuse is open by default.
Does LangFuse only work with LangChain?
No. The naming is historical. LangFuse instruments any LLM stack via SDKs in Python, TypeScript, and OpenTelemetry.
Where is LangFuse data stored?
EU region for the cloud tier by default, with US region available. Self-hosted deployment puts the data wherever your Postgres runs.
