DEV Community

Anmol Gupta
Anmol Gupta

Posted on Originally published at testmuai.com

Let your coding agent prove its UI changes work: Kane CLI in 5 minutes

#testing #ai #cli #webdev

Coding agents can build a feature from a prompt, but they can't tell you whether it actually works in a real browser. Kane CLI closes that loop: you describe a user flow in plain English, it runs it, and it returns an evidence-backed pass or fail.

Disclosure: I work at TestMu AI, which builds Kane CLI.

Install 

npm install -g @testmuai/kane-cli      # Node.js 18+
# or: brew install LambdaTest/kane/kane-cli
kane-cli login
Enter fullscreen mode Exit fullscreen mode

Run your first check 

kane-cli run --url https://example.com "Click the 'More information' link and verify the page loads"
Enter fullscreen mode Exit fullscreen mode

No selectors, scripts or test framework. Every run saves steps, screenshots, console and network logs, and the verdict as an evidence file.

Hook it into your coding agent

  • Claude Code, Codex CLI, Gemini CLI: npx @testmuai/kane-cli-skill
  • Cursor, Copilot: point the agent to AGENTS.md
  • Run with --agent --headless to get structured NDJSON output. Exit codes: 0 pass, 1 fail, 2 setup/auth error, 3 timeout, so the agent (or CI) can gate on them.

Where it runs

Web in real Chrome, API, Android Emulator, iOS Simulator and real devices. It runs headless in GitHub Actions, GitLab CI, Jenkins and Bitbucket Pipelines, and any recorded flow exports to Playwright code.

It's free to install; runs use KaneAI credits, and new users get free credits to start.

Repo: https://github.com/LambdaTest/kane-cli · Docs: https://www.testmuai.com/docs/kane-cli-introduction

How are you verifying agent-written code today?

Top comments (0)