Zurich-based start-up Klarent, formerly Fore AI, has announced an $8 million seed funding round. Its stated goal: to establish itself in the United States with its autonomous software testing platform, after growing in Europe. The information comes from a press release published on 12 October 2026 by Startupticker.ch; the customer references and performance described below are those claimed by the company and have not been independently verified.
Two former Googlers at the helm
Klarent was founded by two former Google engineers: Asheem Panakkat, chief executive officer, and Momchil Ivanov, chief technology officer. The Zurich-based company wants to help businesses improve the quality of their software, “especially in the era of AI-generated software”, according to the press release.
The round is led by Mosaic Ventures, with participation from Moonfire Ventures and a group of angel investors, including Jürgen Galler (a former director at Google), Harris Barton and AngelInvest Ventures.
How agent-driven testing works
To understand the approach, one has to start from the field it targets: quality assurance, or QA. Before putting an application into production, teams write tests, that is, scripts that simulate actions (logging in, filling a shopping cart, paying) and check that the result obtained matches the expected result. Writing and above all maintaining these scripts takes time: every change to the interface or to a process can break some of them.
A test “agent”, in Klarent’s sense, is a system that relies on a large language model (LLM) to take over part of this work. The distinction claimed by the company is important: its agents use LLMs solely to write and adapt so-called “deterministic” test code. In other words, the AI writes the test, but it is not the AI that runs it. Once produced, the script behaves like a conventional test: under the same conditions, it returns the same verdict on every run.
This choice addresses a known limitation of language models: their answers can vary from one prompt to the next. A test whose verdict changed for no apparent reason would be useless for deciding whether a piece of software can be shipped. By confining AI to the writing phase, Klarent says it combines the speed of AI with the precision and cost control of production engineering. The press release does not, however, detail the method used and publishes no comparative reliability measurements.
A human in the loop
The second element of the system: the platform is backed by a human verification service (“human-in-the-loop”) that checks every result. According to the company, this mechanism allows customers to pay only for results. This human involvement also indicates that the automation is not complete: final validation remains, at least in part, a matter for people.
This is one of the limitations to bear in mind when faced with talk of “autonomous” testing. A test, even one written by an AI, only checks what it is asked to check. The question of coverage, that is, which scenarios were tested and which were overlooked, remains open, and the press release does not address it.
Customers in Europe, a US target
Klarent claims rapid adoption in Europe, with customers in several sectors, including JD Sports in e-commerce, the NZZ in media and Sixt in travel and mobility. At JD Sports, its agents reportedly run journeys that cut across several systems in the e-commerce chain, something that even the in-house teams specialised in testing (SDETs, for “software development engineer in test”) had not been able to reproduce, according to the company. The company also says it is expanding into financial services, with leading banks and insurers that it does not name.
The funds will be used to establish a presence in the United States, expand the engineering team and deepen the cross-platform product, on the web as well as on mobile.
A bet on a bottleneck
The lead investor’s thesis is set out by Chandar Lal, partner at Mosaic Ventures: “With AI, code generation is advancing faster than our ability to test it, so QA is going to become the main bottleneck.” He says he was convinced by the founders’ insistence on determinism, namely “using AI to write the tests, not to run them” (quotes translated back from our French version of the English original).
This is a market hypothesis, not an established result. It rests on a plausible trend, more code produced with the help of AI assistants and therefore more code to check, whose scale remains to be measured. For Klarent, the challenge will be to show that its approach delivers on its promises outside Europe and at a larger scale.
