TestMu AI Introduces Source-to-Verdict Loop in Kane CLI, Carrying All Needs to Ship Verdict with Physical Evidence.

SAN FRANCISCO & NOIDA, India — TestMu AI (formerly LambdaTest), the world's first Agentic AI-powered Quality Engineering platform, has introduced a source-to-decision loop in Kane CLI, its natural language testing tool. Building on Kane CLI's evolution from the browser automation tool, the loop carries the product requirement to the ship decision, writing the tests, running them in a local browser, collecting evidence, measuring the coverage of the real event, and returning the decision.

As AI agents write, run, and modify tests, a green check mark is no longer enough: a step that doesn't check that nothing passes, a test written against a feature that changed in the past weeks passes, and an agent that says “done” without a click ever comes and goes. Blank green and earned green are the same color throughout the dashboard, the number goes up while the risk doesn't move. Kane CLI closes that gap: it carries the requirement to the decision, and proves all its steps.

You no longer provide Kane CLI with a test script, you provide it for your needs. Nine sections form one traceable line:

The first part is AI that performs test-engineering; the second part is evidence. The number is finally read from the machine, never written, never guessed.

• Import anything. Point Kane CLI at a PRD, Jira ticket, Confluence spec, Figma frame, or demo video and it pulls them into context – no other tool takes a Figma node or mp4 as a requirement source.
• AI designs tests, and asks when it's due. It extracts business use cases, breaks each one down into scenarios, and pins every scenario to acceptance criteria. If the two sources do not agree, they stop and ask which is the current, rather than guessing.
• The tests you can read, carry their own map. The output is test.md: Clear, editable Markdown, where each step indicates what standard it covers. Traceability is built in, not a spreadsheet on the side of the test.
• Customization in the original browser. Writing is free and objective; execution replays the recorded run without LLM in the loop, so it behaves the same way every time. Automatic cooling comes in only when the product has been flooded, and the flow can exceed 50 steps.
• One packet .and proof per run. Not the screenshot and the log, but the whole system: the trajectory of the agent, the network as HAR, the DOM where things broke, the console, the video, and the documentation of the cause of any failure. It opens as a page with standards – L0 minimum, L1 coverage and quality signals, L3 signing, authentication, and hashes for each file.
• Three points after every marker. Evidence is supported (it was a real passport), to be legal (did the agent actually do it – offline clicks are readable legal:false), and determination (regenerates).
• Number of cover allowed to fall. Coverage is calculated from the package, not the device that used it. Hard mode drops tests past spec from the switch – the percentage that drops is the only type you can trust if it goes up.
• Then someone signs. Coverage, confidence, non-covered conditions, and old sources come to a single decision, ship or hold, posted as a required check on the pull request, with a link to the proof attached. The machine brings the evidence; someone is calling.

The package is not locked inside the dashboard, is plain YAML and Markdown, versioned, git-distributable, and readable by humans, agents, and auditors. Teams can take it without adopting the Kane CLI itself, and package the project in CI formats like CTRF and JUnit. Think of it as a PDF of the test proof.

TestMu AI (formerly LambdaTest) is the world's first Agenttic AI-powered Quality Engineering platform, helping teams build, test, and release software with confidence in the AI-first era by combining the power of autonomous testing with real-world validation. For more information, visit www.testmuai.com.

https://www.businesswire.com/news/home/20260723455596/en/

Corporate Communications Manager