Langfuse

Open-source LLM observability platform providing comprehensive tracing, analytics, and evaluation for AI applications with self-hosting capability and a generous free tier. Strong open-source traction; 50K+ events on free tier; integrates with LangChain, LlamaIndex, OpenAI, and Anthropic SDKs; YC-backed; growing European developer community; Core plan at $29/mo.

Langfuse's open-source model and self-hosting capability make it the preferred observability layer for privacy-conscious teams deploying AI coding agents, competing directly with LangSmith.

langfuse.com Observability and AI Monitoring

Germanyprivate

At a glance

Total raised

$4.5M

last round acquired

Headcount

11-50

approximate · public range

Headquarters

Berlin

Germany

Founded

2022

4 years active

Status

Private

Observability and AI Monitoring

Subsector

Observability and AI Monitoring

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Founding

50% off

/month

94 of 100 spots left

Full

$10

/month

Available once all 100 Founding Member spots are claimed.

Get full access

Already subscribed? Sign in →

Funding history

Total raised

$4.5M

Rounds tracked

Latest stage

Acquired

Date	Stage	Amount	Valuation	Lead investors
Jan 16, 2026	Acquired	—	—	ClickHouse
Nov 7, 2023	Seed	$4M	—	Lightspeed Venture Partners, La Famiglia
Nov 1, 2022	Convertible	$500K	—	Y Combinator

Founder backgrounds

Clemens Rawert

Co-Founder & COO

Before co-founding Langfuse, worked in the CEO Office and Investment Management at Scalable Capital, a German fintech unicorn, where he collaborated closely with the founder-CEOs on strategy and operations. He holds degrees from the London School of Economics and Political Science (LSE) and the University of Oxford. Langfuse, his second venture, was part of Y Combinator's W23 batch and provides an open-source observability platform for LLM applications, supporting self-hosting and a generous free tier.

Marc Klingen

Co-Founder & CEO

Holds a Master's in Management & Computer Science from the Technical University of Munich, graduating in the top 1% of his class, and participated in the Center for Digital Technology and Management (CDTM). Prior to founding Langfuse, he has described his background as moving from traditional business roles into engineering and entrepreneurship. Langfuse was founded in 2022, participated in Y Combinator's W23 batch, and is built on ClickHouse — a relationship reflected in his "part of @clickhousedb" descriptor. The company's team page notes it is composed of "experienced engineers and ex-founders." Publicly, he emphasizes a revenue-funded, community-driven approach to building open-source infrastructure, and has spoken about the importance of fair compensation and transparent hiring as core to Langfuse's culture.

Max Deichmann

Co-Founder & CTO

Co-Founder & CTO of Langfuse (YC W23), the open-source LLM observability platform. Prior to Langfuse, built a background spanning both management and computer science, with a focus on building reliable, scalable systems. He led the technical architecture behind Langfuse's tracing and analytics engine, which processes billions of observability events per month for agentic LLM applications — a product he described as "observability for engineering teams working on the most complex and advanced agentic systems." In 2024, Langfuse joined ClickHouse, pairing the platform with ClickHouse Cloud to cut query latency and improve data retention efficiency. He is based in Berlin.

AI bull / bear

Bull case

The ClickHouse acquisition transforms Langfuse from a scrappy YC-backed open-source project into a platform with serious infrastructure muscle. ClickHouse's real-time analytics engine is purpose-built for the high-cardinality, high-volume trace data that LLM observability generates — marrying the two creates a technically formidable stack that Datadog and New Relic cannot easily replicate without rearchitecting their storage layers. Self-hosting remains Langfuse's killer differentiator. Privacy-conscious enterprises in finance, defense, and healthcare — precisely the sectors pouring money into AI coding agents and internal LLM tools — will not ship trace data to LangSmith's cloud. Langfuse gives them the same observability surface area without the data-exfiltration risk. The generous 50K-event free tier continues to seed bottoms-up adoption inside engineering teams, and the $29/month Core plan is priced to convert those teams without procurement friction. European data-sovereignty requirements under the EU AI Act further tilt the playing field toward a platform born in Berlin, not San Francisco.

Bear case

An acquisition by a database company rarely ends with the acquired product flourishing as an independent platform — it ends with the product becoming a feature. ClickHouse's incentive is to use Langfuse as a reference workload that drives ClickHouse Cloud adoption, not to build a generational observability company. If Langfuse gets folded into ClickHouse's go-to-market, the open-source community that fueled its early traction may fragment, especially if self-hosting receives second-class support relative to the managed ClickHouse Cloud path. LangSmith enjoys an intrinsic distribution advantage from being the default observability layer inside the LangChain ecosystem. For teams already building on LangChain, switching to Langfuse means changing instrumentation code — a friction point that kills adoption in production environments. The observability market is also attracting heavyweight entrants: Datadog, Grafana, and cloud providers all have LLM monitoring on their roadmaps, and they bring existing customer relationships and nine-figure R&D budgets. Langfuse's $5 million in total funding, even with ClickHouse's backing, may not be enough to outrun converging competition in a market where the big players can afford to give observability away for free to protect their core platforms.

Generated by deepseek-v4-pro · May 2, 2026Refreshes quarterly · see disclosures

In the news

Langfuse · last 30 days

No articles ingested yet for Langfuse. Once the hourly news pipeline is live, every article the classifier tags as mentioning this company appears here with its one-line AI summary and sentiment.

Watch

Leadership

Marc Klingen

Co-Founder & CEO

Max Deichmann

Co-Founder & CTO

Clemens Rawert

Co-Founder & COO

—

Products

Langfuse Tracing
An open-source observability platform that captures detailed traces of every LLM call, including latency, token usage, cost, and full input/output payloads. Supports any LLM provider or framework via Python and JS/TS SDKs, native integrations with OpenAI, LangChain, LlamaIndex, and Vercel AI SDK, and an OpenAPI-compliant API. Developers use it to debug production issues, monitor performance, and understand user behavior across chains and agents.
Prompt Management
A version-controlled prompt management system that lets teams collaboratively create, iterate, and deploy LLM prompts with full change history, rollback capabilities, and environment-specific configurations (dev/staging/production). Integrated directly into the tracing workflow, it enables A/B testing of prompt variants and automatic linking of prompt versions to the traces they produced, giving teams a complete audit trail.
Langfuse Evaluations
A framework for running automated evaluations on LLM outputs, supporting both LLM-as-a-judge scoring and custom metric functions. Teams can define evaluation criteria, run them on datasets or live traces, and track results over time to catch regressions. Integrates with the tracing pipeline so every evaluation result is automatically linked to the original trace, making it straightforward to correlate changes in prompt or model behavior with quality scores.
Langfuse Datasets
A dataset management tool for curating, versioning, and exporting training or evaluation data extracted from production traces. Teams can create datasets from specific trace patterns, annotate them, and use them to run eval suites or fine-tune models. Built directly on top of the tracing infrastructure, it eliminates the need to manually export and reformat data from logs.

Open roles

View all careers ↗