Member-only story
We crossed 100K traces a month in March. That’s the point where Langfuse Cloud’s Pro tier stops feeling like a rounding error and starts showing up in the “why is our tooling spend up 40%” Slack thread.
Two options. Negotiate an enterprise contract for a tool we mostly used to answer “why did the agent say that.” Or run the exact same open-source stack ourselves, on infrastructure we already pay for.
I took option two. It runs an afternoon to get something working and about a week to get something I’d trust in production. This is the guide I wanted going in.
One thing before we start. Langfuse was acquired by ClickHouse in January 2026. The core platform remains MIT licensed and fully self-hostable, and self-hosting via Docker Compose or Kubernetes is unchanged and continues to be actively maintained. Nothing here depends on a vendor relationship that could evaporate. If you’re picking observability infrastructure for a company that plans to exist in five years, that’s not a footnote.
Why bother, when you could just pay for it #
Langfuse is an open-source LLM engineering platform: tracing, evals, metrics, prompt management, playground, and datasets, instrumented through…