Traceloop raises $6.1 million in Seed funding to stop AI agents from going rogue
The Israel startup offers observability and testing tools to prevent generative AI failures in production.
As the race to integrate generative AI intensifies, one of the industry’s quieter but most pressing problems is now getting its own dedicated fix: how to ensure AI agents don’t go off the rails after they’re deployed. Traceloop, a startup founded by machine learning and enterprise software veterans, announced Tuesday that it has launched its commercial platform and raised $6.1 million in Seed funding to make AI agents production-ready.
Full list of Israeli high-tech funding rounds in 2025
The funding round was led by Sorenson Capital and Ibex Investors, with participation from Y Combinator, Samsung NEXT, and Grand Ventures.
Traceloop’s core mission is to bring engineering discipline to the still-fragile world of AI agents, those software entities that use large language models (LLMs) to act autonomously on behalf of users. While developers have been able to build increasingly complex agents using tools from OpenAI, Google, and Anthropic, they still struggle to track how those agents behave once deployed. Users rarely report bugs. They just stop using the product.
“Prompt engineering shouldn’t be a guessing game or rely on vibes,” said CEO and co-founder Nir Gazit. “It should be like the rest of engineering – observable, testable, and reliable. When we bring the same rigor to AI that we expect from the rest of our stack, we unlock its full potential.”
Related articles:
Gazit and CTO Gal Kleinman, who previously led ML infrastructure at Fiverr, built Traceloop after observing a persistent gap between AI performance in demos and behavior in production. Gazit previously spent four years at Google building LLM-based systems to predict user engagement and retention.
Traceloop’s solution is built on top of its open-source technology, OpenLLMetry, which is now commercially available. The platform replaces manual trial-and-error testing with automated evaluations of agent performance, enabling developers to detect problems early, track real-world behavior, and deploy fixes with greater confidence.
Companies like IBM, Cisco, Dynatrace, and Miro are already using Traceloop’s system to monitor and verify their AI agents. The platform’s open-source foundation has seen rapid adoption: OpenLLMetry counts over 500,000 monthly installs, with 50,000 weekly active SDK users and more than 60 contributors on GitHub.
