langfuse ClickHouse Langfuse joins ClickHouse Our goal continues to be building the best LLM engineering platform Read story

Langfuse joins ClickHouse Our goal continues to be building the best LLM engineering platform ClickHouse has acquired Langfuse. If you're reading this as a Langfuse user, your first question is probably: What does this mean for me? Our roadmap stays the same, our goal continues to be building the best LLM engineering platform, and we remain committed to open source and self-hosting. There are no immediate changes to how you use Langfuse and how you can reach out to us. What does change is our ability to move faster. With ClickHouse behind us, we can invest more deeply into performance, reliability, and our roadmap that helps teams build and improve AI applications in production. What stays the same This is the section we would want to read first, too. - Langfuse stays open source and self‑hostable. There are no planned changes to licensing. As you know, we leaned heavily into OSS over the last years. - Langfuse Cloud keeps running as‑is. Same product, same endpoints, same experience. - Support stays the same. Same channels, same SLAs for existing customers. What gets better now Joining Clickhouse compresses years of operational learning into immediate, real customer benefits. - More engineering leverage on the hardest parts. Langfuse is a data‑intensive product. Working closely with the ClickHouse engineering team helps us push performance and reliability. - Faster progress on enhanced enterprise-grade compliance and security, with the help of Clickhouse's resources. - Learning from Clickhouse's customer success and support playbook. This puts us years ahead and allows us to spend more time on what we really care about: our users. A quick look back The longer version of how we got here is in our handbook. Langfuse started the same way many LLM products start: we were building agents ourselves. And we constantly ran into the same problems. Building LLM apps is easy to demo and hard to run in production. Debugging is different, quality is non‑deterministic, and the iteration loop is messy. When we did Y Combinator in early 2023, we saw this every week, both in our own projects and in what other founders in our cohort were working on. So we built a duct tape version of what we wished existed: tracing and evaluation primitives that are easy to add, easy to self‑host, and actually useful for iterating. The very first version was intentionally simple. It ran on Postgres, because speed of shipping mattered more than theoretical scaling. That got us to a real product and a real community fast. Then people actually started to use the product more than we could have imagined. As adoption grew, Postgres became the bottleneck for the workloads Langfuse needed to support (high‑throughput ingestion + fast analytical reads). With Langfuse v3, we switched the core data layer to ClickHouse to make Langfuse scale for production workloads, both in Cloud and self‑hosted deployments. And if you like infrastructure deep dives, here’s the v3 migration write‑up. Why join ClickHouse There are a lot of ways this could have…