Deals · Enterprise AI

Klarent raises €7.13M Seed to scale agentic AI software testing and expand into the US

A Zurich-based agentic software testing platform that uses LLMs to write test code for web and mobile applications, then runs that code deterministically — with a human-in-the-loop layer and outcome-based pricing — for enterprise customers who cannot keep pace with AI-accelerated code generation.

ProYarn Desk · October 9, 2026Read this in French

in 𝕏 ✆

“Code generation is outpacing testing capacity and QA will become the main bottleneck.”

Chandar Lal — Partner, Mosaic Ventures

Klarent (Zurich) has raised €7.13 million (~$8.1M) in a Seed round led by Mosaic Ventures (London), with Moonfire Ventures participating, alongside angel investors including Jürgen Galler (former Google director) and Harris Barton. The company, which previously operated as "fore ai", builds an agentic software testing platform for enterprise web and mobile applications.

The architecture: AI writes, determinism runs

Software testing has two expensive steps: writing test code and running it. The first step requires understanding application logic, user flows, and edge cases — work that historically consumed QA engineers' time and was bottlenecked by their availability. The second step is computationally intensive but, once tests are written, can be automated.

Klarent's architecture applies AI to only the first step. LLMs generate the test code — analysing application surfaces, writing assertions, structuring test suites — while execution remains deterministic. A human-in-the-loop layer reviews generated tests before they run in production. Customers are billed for verified outcomes, not for test iterations.

The design choice matters: fully AI-generated test assertions are unreliable at enterprise scale because LLMs can assert behaviour that sounds correct without verifying it against the actual application. By separating authoring (AI) from execution (deterministic), Klarent sidesteps the hallucination failure mode while still capturing speed gains at the bottleneck step.

Enterprise traction and the US expansion

Current enterprise customers — JD Sports (UK retail), NZZ (Swiss media group), and Sixt (German mobility) — span industries and geographies. The €7.13M will fund US market entry and vertical expansion into financial services and mobile applications.

"Code generation is outpacing testing capacity and QA will become the main bottleneck," said Chandar Lal, partner at Mosaic Ventures. The thesis is structural: AI coding tools raise developer throughput, but QA capacity hasn't kept pace — the verification gap grows as the code-generation gap closes.

What the expansion needs to prove

The US market brings established QA toolchain preferences and conservative enterprise procurement. Klarent will need to make its case at the engineering leadership level — where the AI throughput argument lands cleanest — rather than with QA teams already embedded in Selenium or Cypress workflows. The next 18 months will establish whether agentic test authoring is a new category, or whether it gets absorbed into the AI-coding platforms that large enterprise software vendors are already building.

Why it matters to investors

The relationship between AI-generated code and software quality has a straightforward problem at its core: the same AI tools that accelerate writing code do not automatically improve the quality of the tests that verify it. Developer throughput has increased; QA throughput has not kept pace; the result is a growing gap between code shipped and code verified. Klarent's architecture addresses this with a specific trade-off: LLMs are used only for the expensive part of the process that has historically required human expertise — writing test code for novel application surfaces — while execution remains deterministic. Customers pay only for verified outcomes, not for test runs. This is a considered design: it avoids the 'hallucinating tests' failure mode that makes fully AI-generated test assertions unreliable, while capturing the speed advantage at the authoring stage. The customers already operating on the platform — **JD Sports** (UK retail), **NZZ** (Swiss media group), and **Sixt** (German mobility) — span three industries and two geographies, which limits the argument that the product only works for a particular enterprise archetype. **Mosaic Ventures** (London), the lead, backs early-stage enterprise software from a broad market perspective; partner Chandar Lal's note that 'code generation is outpacing testing capacity and QA will become the main bottleneck' is a clean statement of the thesis: not that AI testing is incrementally useful, but that QA is structurally underinvested relative to the pace at which AI is generating code that needs testing. **Moonfire Ventures** adds complementary coverage of the European-to-US expansion trajectory. The US expansion funded by this round is the real growth test: Klarent's customer base is currently European, and enterprise QA procurement in the US has well-established toolchain preferences (Selenium, Cypress, TestRail, established vendors). Entering as an agentic AI-first platform — not as a drop-in replacement for existing tools — requires a buying argument that is compelling at the CTO or VP-Engineering level rather than the QA team level, where incumbent tools are already embedded.

Sources

Seed Enterprise AI Switzerland Mosaic Ventures Moonfire Ventures

Threaded to this story

Enterprise AI · September 2, 2026

Every European round, in your inbox by 8am.

The day's seed and Series A rounds across France and Europe — threaded, sourced, and read in two minutes. Free.