Marc Klingen is a cofounder of Langfuse and leads its team at ClickHouse. Langfuse is an open-source platform for tracing, evaluating and improving applications built with language models and agents.[1]
Start with Langfuse →
Education
Links
WebsiteLead, Langfuse team · ClickHouse
Cofounder · Langfuse
Career
Klingen cofounded Langfuse with Max Deichmann and Clemens Rawert during Y Combinator’s W23 batch. His personal account says he joined ClickHouse in January 2026 and now leads the Langfuse team. Before Langfuse, he worked at Google, DHL, Cherry Ventures and startups in engineering, product, business intelligence and sales. He studied business as an undergraduate, then computer science and innovation at the Technical University of Munich and joined CDTM.[1]
Contributions
Langfuse records the steps and model calls in an application so teams can inspect what happened during an agent run. It connects those traces with evaluation, datasets and prompt management. The founders’ acquisition announcement describes the goal as shortening the loop from production observations to better prompts, tests and product decisions. The contribution is a shared engineering workflow for understanding and improving model behavior.[2][3]
Ideas
The ClickHouse announcement connects Langfuse’s future work to production monitoring, analytics and large deployments. It also emphasizes maintaining its open-source identity and developer experience. These priorities address a practical tension: detailed traces help explain failures, but collecting and querying them must stay workable as an application produces more runs and a team conducts more experiments.[2]
Tools, writing & talks
Tools
Start here
Langfuse
github.comAn open-source platform combining tracing, evaluation, prompt management and datasets for language-model applications.
Start at the repository to see the tracing and evaluation workflow together, then follow the quickstart for your language. A small existing language-model application gives you real runs to inspect.
Writing, research & talks
Langfuse's co-founder walks through the online and offline loops for improving agents, covering traces, datasets, experiments and evals. A demo has a coding agent find jargon leaks, update evals and back-test a fix, while humans keep goal-setting.
Catalog workOriginal source- Langfuse joins ClickHouselangfuse.com
The founders explain their acquisition, engineering priorities and plans for production evaluation workflows.
Coauthor with fellow founders[2]