ClickHouse, the company behind the open-source analytical database of the same name, acquired Langfuse, an open-source platform that developers use to trace, evaluate and monitor applications built on large language models. Both companies announced the deal on 16 January 2026, and the whole Langfuse team joined ClickHouse.
Neither side disclosed the price or how it was paid. ClickHouse announced the acquisition on the same day it said it had closed a $400 million Series D round led by Dragoneer Investment Group. Orrick, Langfuse's law firm, said that round valued ClickHouse at $15 billion.
Langfuse stays open source under its existing MIT licence and can still be self-hosted. Langfuse Cloud keeps running as a separate service, existing customers keep their support channels and service levels, and the founders said the product roadmap had not changed. ClickHouse said it planned deeper integrations over the following months to make LLM observability part of what it calls its Agentic Data Stack.
What Langfuse built
Langfuse was founded in 2023 by Marc Klingen, Max Deichmann and Clemens Rawert. The founders had been building LLM agents themselves and kept running into the same problems once those agents reached production. They went through Y Combinator's Winter 2023 batch.
The product records each interaction with a model as a trace, scores those interactions against quality measures such as hallucination checks, and tracks the results over time. It also covers prompt management, evaluations, experiments, and cost and latency tracking, and it supports open standards such as OpenTelemetry.
According to ClickHouse, Langfuse had more than 20,000 stars on GitHub, more than 23.1 million SDK installs a month and more than 6 million Docker pulls. ClickHouse said 19 of the Fortune 50 and 63 of the Fortune 500 used it, naming Intuit, Twilio, 7-Eleven and Merck among enterprise users. Orrick said the company had more than 2,000 paying customers.
“We built Langfuse on ClickHouse because LLM observability and evaluation is fundamentally a data problem.”
Already running on ClickHouse
The two companies were linked technically well before the deal. Langfuse's first version stored its data in Postgres. With version 3, the core data layer moved to ClickHouse, and both the cloud service and self-hosted installations now run on it. The founders wrote that Langfuse Cloud was a large ClickHouse Cloud customer and that the move from version 2 to version 3 introduced thousands of teams to ClickHouse.
The relationship also ran the other way. ClickHouse said it used Langfuse internally to monitor DWAINE, an AI agent that works on its internal data warehouse. The founders described the acquisition as a way to make an existing partnership permanent and to invest in it more heavily.
Why ClickHouse is buying
ClickHouse argued that companies building generative AI applications need to see how those applications behave in production before they can trust them. Klingen, Langfuse's chief executive, said the company had built on ClickHouse because observing and evaluating LLMs is at heart a data problem.
Analysts quoted by InfoWorld read the deal in a similar way. Sanchit Vir Gogia of Greyhound Research said it widened ClickHouse's audience to AI engineering and compliance teams. Ishi Thakur of Everest Group said data platform vendors were moving up the stack, from systems that store and analyse data towards systems that take action.
The purchase was part of a wider expansion announced the same day. ClickHouse also launched a managed Postgres service built with Ubicloud, and said it had more than 3,000 customers on ClickHouse Cloud.
The founders and the funding
Langfuse raised a $4 million seed round in November 2023 from Lightspeed Venture Partners, La Famiglia and Y Combinator, according to its own announcement. Orrick and InfoWorld also list General Catalyst among its backers. InfoWorld reported that Langfuse had a team of 13, with an office in San Francisco for marketing and sales alongside its Berlin base.
The founders wrote that they had not planned to sell and had term sheets for a Series A round when they agreed to the deal. They said joining ClickHouse would bring more engineers for performance and reliability work, faster progress on enterprise compliance and security, and access to ClickHouse's customer support teams.
A crowded field for AI observability
ClickHouse is not the only data platform moving into the area. InfoWorld pointed to Databricks, which offers its own observability features, and to Snowflake's acquisition of Observe as a similar step towards built-in observability.
ClickHouse's Series D was led by Dragoneer, with Bessemer Venture Partners, GIC, Index Ventures, Khosla Ventures, Lightspeed Venture Partners, accounts advised by T. Rowe Price Associates and WCM Investment Management also taking part.
