Klarent, a Zurich, Switzerland startup whose software writes and maintains the tests that companies run before releasing an app, has raised €7.13M, or about $8M, in a seed round. London’s Mosaic Ventures led the round, joined by Moonfire Ventures, AngelInvest Ventures and angel investors including Jürgen Galler, a former Google director, and the former American football player Harris Barton.
The company, which traded as fore ai until its rename, plans to expand into the United States (US) and into financial services.
“With AI, code generation is scaling faster than our ability to test it, so QA will become the key bottleneck,” said Chandar Lal, a partner at Mosaic Ventures. QA, short for quality assurance, is the work of checking that software behaves as it should.
AI Writes The Tests, Not The Verdict
Every time a shop updates its checkout page or a bank changes its app, someone has to confirm that the basic journeys still work: logging in, searching, paying, getting a receipt. Large companies run thousands of these checks, and much of the effort goes into keeping old tests working after the screens they rely on have changed. Klarent says these regression checks, which re-run old tests after every update, typically take up 40% to 50% of a testing team’s time.
Klarent hands that job to a set of AI agents, programs that carry out a task in steps without a person directing each one. An Explorer agent maps the screens and paths through an app, a Planner designs test cases, a Verifier checks the plan, a Coder writes the test scripts and a Runner executes them, then reports the results. Teams can start from a web address, a product document, a Jira project-management ticket or a design file, or simply describe in plain English what they want checked.
Large language models (LLMs), the technology behind chatbots, write and adapt the test code in Klarent’s system, but they do not run the tests. Once written, each test runs the same way every time, which the company says avoids the unpredictable results AI can produce and keeps the cost of running AI down.
When an app’s screens change, the platform proposes fixes to the broken tests and shows each change step by step for a person to approve. The platform also plugs into the services developers use to store and release code, such as GitHub and GitLab, so tests can run automatically whenever new code is added. It sends results to workplace chat apps such as Slack or Microsoft Teams.
A human verification layer sits on top, with people checking the results, and Klarent says customers pay only for verified outcomes. Its website describes annual licences that scale with the number of tests generated and run. Asheem Panakkat, Klarent’s chief executive, said that “testing stops being a cost centre and becomes the thing that determines whether any of that code can be trusted”.
Where It Is Already Working
At the sportswear retailer JD Sports, Klarent’s agents automate checks that cross several systems in the retailer’s online shopping pipeline, work that JD Sports’ own test engineers had not been able to replicate. The car rental group Sixt is also a customer.
The Swiss newspaper group NZZ is the most detailed example on Klarent’s own website. Klarent says the publisher’s quick checks after each release now run 98% faster, and that product and business staff rather than engineers create 75% of the publisher’s tests.
“We shifted our perspective on how we do QA,” said Volker Dietzel, NZZ’s chief technology and information officer.
NZZ’s QA team lead, Pieter Wiesner, said the team can now check paywalls that differ by country and schedule tests to run every 15 or 30 minutes. The fuel retailer OnlineFuels runs more than 130 automated tests that finish in under four minutes, and its chief executive, Marc Noy, has said the biggest gain was turning manual test cases into automated ones. Klarent also says an unnamed large bank has cut its manual testing effort by about 90%.
The website also lists the Swiss bank UBS, the ride-hailing company Uber and Google among those using the platform. It says the software can run in Klarent’s cloud, in a private cloud or on a customer’s own servers.
Why Testing Became The Bottleneck
Panakkat spent more than a decade at Google, leading teams on its Shopping and Lens products, while Klarent’s technology chief, Momchil Ivanov, was a staff software engineer there working on information retrieval. They founded the company in 2023.
AI coding assistants and agents, which write software from a plain description, can turn out working features in minutes. Each change still has to be tested before it reaches customers, and mistakes slip through when it is not. Morning Tick has reported on a case where AI coding agents published 13,000 private screenshots, and on startups such as Rein raising money to keep AI agents from acting outside their limits. Lal said the founders had spotted the testing problem at Google before most of the market, and that their insistence on tests that run the same way every time was a major reason Mosaic invested.
Next Stop: The US
Klarent plans to use its European enterprise customer base to enter the US, grow its engineering team and extend its testing across web and mobile. It also plans to push into financial services, where it says it is already working with large banks and insurers. It has not disclosed its revenue, the size of its team or its valuation, nor where in the US it will base its first staff.
The market analyst CB Insights compares Klarent with the AI testing startups Octomind and ProdPerfect. Klarent is betting that companies will want predictable, human-confirmed testing as more of the software they ship is written by machines.
