Ravenna agents automate the IT service desk of the Langfuse and ClickHouse teams! To make sure those agents work as expected, they observe and improve them with Langfuse. The favorite request in the Langfuse team: get an email alias with your first name assigned by quickly asking in the IT support channel. Check the full write up on our blog 👇
Langfuse
Software Development
Open Source LLM Engineering Platform, now part of ClickHouse
About us
Langfuse is an open source AI engineering platform. It helps teams collaboratively develop, monitor, evaluate, and debug AI applications. Langfuse can be self-hosted in minutes and is battle-tested and used in production by thousands of users from YC startups to large companies like Khan Academy or Twilio. Langfuse builds on a proven track record of reliability and performance. Developers can trace any Large Language model or framework using our SDKs for Python and JS/TS, our open API or our native integrations (OpenAI, Langchain, Llama-Index, Vercel AI SDK). Beyond tracing, developers use Langfuse Prompt Management, its open APIs, and testing and evaluation pipelines to improve the quality of their applications. Product managers can analyze, evaluate, and debug AI products by accessing detailed metrics on costs, latencies, and user feedback in the Langfuse Dashboard. They can bring humans in the loop by setting up annotation workflows for human labelers to score their application. Langfuse can also be used to monitor security risks through security framework and evaluation pipelines. Langfuse enables non-technical team members to iterate on prompts and model configurations directly within the Langfuse UI or use the Langfuse Playground for fast prompt testing. Langfuse is open source and we are proud to have a fantastic community on GitHub and Discord that provides help and feedback. Do get in touch with us! Langfuse is now part of ClickHouse.
- Website
-
https://langfuse.com
External link for Langfuse
- Industry
- Software Development
- Company size
- 11-50 employees
- Headquarters
- San Francisco
- Type
- Privately Held
- Founded
- 2022
- Specialties
- Langfuse, Large Language Models, Observability, Prompt Management, Evaluations, Testing, Open Source, LLM, AI, Analytics, Open Source, and Artificial Intelligence
Products
![Click here to view Langfuse]()
Langfuse
Product Analytics Software
Langfuse is an open-source LLM engineering platform that helps teams collaboratively debug, analyze, and iterate on their LLM applications.
Locations
-
Primary
Get directions
San Francisco, US
-
Get directions
Berlin, DE
Employees at Langfuse
Updates
-
The Agentic Data Stack is coming to Berlin on September 2nd. Join us for an evening that is all wbout what it takes to run agents in production. Happy to co-host this event with ClickHouse, and Merantix Momentum! Link for sign-ups in comments 👇️
Demos are easy. Production is humbling.🤖 On 2nd September, ClickHouse, Langfuse and Merantix Momentum are hosting an evening at the Merantix AI Campus in Berlin. 🇩🇪 Theme: What it actually takes to run agents in production. Talking to teams across different startups all year, the same four things keep breaking: 1️⃣ Traces nobody can query 2️⃣ Evals that live in someone's head 3️⃣ Models never tuned to the job they do 4️⃣ A data layer built for dashboards, not agents That is the whole agenda for the evening: 🎤 Bob De Vries (Solutions Architect, ClickHouse) on the data backend for agentic workloads 🎤 Annabell Schäfer (Growth Engineer, Langfuse) on observability and evals for LLM apps Then a panel with people who have shipped it and lived with the consequences: 🌟 Dustin Gedlich (Principal Engineer, Peec AI), 🌟 Selim Nowicki (Co-Founder, distil labs), and 🌟 Marcel Weigand (Sr. AI Solutions Architect, Merantix Momentum), 🌟 moderated by Valeriia Dragan. No pitch decks. Just engineers being honest about what failed before it worked. If you are building agents and want to compare notes over a drink, send me a DM to save you a spot. Invite only, free registration: https://lnkd.in/eaZRtSpu #AgenticAI #LLMOps #AI #Observability #Berlin #TheAgenticDataStack #Startups #ClickHouse #Langfuse #PeecAI #distillabs #Merantix #Data
-
-
𝗟𝗮𝗻𝗴𝗳𝘂𝘀𝗲 𝘃𝟰 𝗶𝘀 𝗹𝗶𝘃𝗲 𝗼𝗻 𝗟𝗮𝗻𝗴𝗳𝘂𝘀𝗲 𝗖𝗹𝗼𝘂𝗱 𝗮𝗻𝗱 𝗚𝗔 𝗳𝗼𝗿 𝘀𝗲𝗹𝗳-𝗵𝗼𝘀𝘁𝗲𝗿𝘀. v4 is the biggest change to the platform since launch: it is a re-architecture of our data model. It is more performant and enables new features - like the new observations table, full-text search, a filter search bar, monitors, code evaluators, and the Langfuse assistant. v4 is live on Langfuse Cloud and generally available for self-hosted deployments. Steffen Schmitz and Hassieb Pakzad walk through everything in the video. Upgrading might require changes to your Langfuse setup (e.g. instrumentation, SDKs, APIs, evals). On Langfuse Cloud, open the v4 Upgrade Check to see which changes apply to your project. Self-hosters can upgrade on their own schedule. 🔗 Link in comments
-
"Building a product like ours would be impossible without Langfuse..." Ravenna builds an AI-native internal service desk where AI and human agents collaborate across IT, HR, and operations. Because its agents navigate hundreds of systems, request types, and company-specific contexts, observability was essential from day one. Today, Langfuse gives Ravenna end-to-end visibility into every agent interaction, with impressive results: ✔️ Agent workflows debugged orders of magnitude faster ✔️ One-click traces connecting TypeScript, Python, Slack, and the web app ✔️ Customer success teams diagnosing issues without pulling in engineers ✔️ Production feedback feeding a continuous agent improvement loop Check out the full story on our blog, link in comments 👇
-
Introducing Leonard Wolters 👋 In the last months, we doubled our team. We nearly forgot to introduce them to you. We’re coming clean now. Leo joined us to jump on any project that needs urgent attention, building the operational backbone of the company. A few things he’s done so far: 🕵️ Owned recruiting and people ops as we doubled the team in his first six months. Thousands of applications, hundreds of interviews, revamped processes to find the best additions to our team. 🌍 Drives field marketing for Langfuse, making sure Langfuse shows up around the world. Organized events with OpenAI, Anthropic, Superhuman, and more (to come). 🧾 Customer billing, payroll, accounting, regular ops things in the background work that keep the company rolling. It's great to work with you, Leo.
-
-
Curious what it's like behind the scenes at Langfuse? Caleb Love Seeling filmed a "Day in the Life" as a Technical Support Engineer on our team. Have a look, and if it looks like your kind of thing, we're hiring!
-
Overview of what's new in 2026. Very productive year so far, more to come!
This year has been really good for Langfuse so far. Quick summary of largest releases from January through July. Want to see me demo everything on the slide and hear what comes next? Join our community town hall next week (link in comments)
-
-
Memory is one layer of the agent stack observability often can't see. EverOS now speaks OTel natively, so memory operations land in Langfuse: what got stored, what got recalled, how confident that recall was, how conflicting memories got merged, and what it cost.
Memory is the one layer of the agent stack observability can't see. That makes memory a hidden decision layer: what gets stored, what gets recalled, what gets consolidated, and what gets forgotten. Every one of those choices shapes what your agent says next, and none of them show up in a log line. When the agent gets it wrong, you are guessing. We build EverOS, an open-source memory engine for AI agents (Apache-2.0, local-first). As of 1.2.0 it emits OpenTelemetry natively, so those decisions stop being hidden. Our Langfuse integration page went live today: https://lnkd.in/gPM68PDD What that makes visible: • 𝗧𝗵𝗲 𝗺𝗲𝗺𝗼𝗿𝘆 𝗹𝗶𝗳𝗲𝗰𝘆𝗰𝗹𝗲 𝗮𝘀 𝗮 𝘁𝗿𝗮𝗰𝗲 𝘁𝗿𝗲𝗲. One write opens a trace, and every stage becomes a step inside it: boundary detection, extraction, reflection, and storage. Searches get a trace of their own, from the query embedding through recall and ranking. Each step carries its own time and cost, so the price of remembering one thing is itemised rather than arriving as a single number at the end of the month. • 𝗖𝗼𝗻𝗳𝗹𝗶𝗰𝘁𝗶𝗻𝗴 𝗺𝗲𝗺𝗼𝗿𝗶𝗲𝘀 𝗴𝗲𝘁𝘁𝗶𝗻𝗴 𝗿𝗲𝘀𝗼𝗹𝘃𝗲𝗱. When a plan changes or a preference is updated, EverOS Reflection merges the two memories and retires the outdated one, which then stops coming back from search. This normally happens invisibly, and it is the hardest thing to debug when it goes wrong. Now it shows up in the trace, so you can see which conversation changed what your agent believes. • 𝗥𝗲𝗰𝗮𝗹𝗹 𝗰𝗼𝗻𝗳𝗶𝗱𝗲𝗻𝗰𝗲 𝗼𝗻 𝗲𝘃𝗲𝗿𝘆 𝘀𝗲𝗮𝗿𝗰𝗵. Each search returns a score for how good EverOS estimates the recall was. Trend it over time and you catch the queries that quietly come back with nothing useful. There is no wrapper library and no instrumentation code to write. Tracing is one config block, off by default, and content capture is opt-in. Stop guessing why your agent forgot. Check the integration page for the full setup, or replay our recording of a real run into your own Langfuse project, no install required. Thanks to the Langfuse team for collaborating on this! 🙌
-
-
Quarterly Langfuse Town Hall. Join the Langfuse team, maintainers, and community for our quarterly call. Marc Klingen will demo the features we shipped in the last 3 months and share more about our upcoming roadmap. Afterward, we'll have lots of time for Q&A. 𝗪𝗵𝗲𝗻: Wednesday, August 12th, 2026, at 9am PT / 6pm CEST 𝗪𝗵𝗲𝗿𝗲: Virtual on Google Meet Link in comments.
-
-
Next stop on our Europe Tour: 🇬🇧 London End of September, we'll be in town for Open House by ClickHouse and host a hands-on workshop "From 0 to Production", showing you in three hours how to bring an agent to life using Langfuse. Sign up in the comments ⬇️
-