Kane CLI by TestMu AI
Natural language in. Real browser. Pass or fail out.
4.0
(11 평점)
AI agent
Browser AI Agent
카테고리
Browser AI Agent
가격 모델
프리미엄 - 기본 무료, 고급 유료
산업 분야
기술
Kane CLI is a browser agent that verifies web applications from natural language. You describe what to test the way you would explain it to a teammate, and Kane CLI carries out the flow in a real Chrome browser, then returns a clear pass or fail with evidence behind it.
It is built for developers who want to check a change before raising a pull request, QA teams who want browser coverage without maintaining a framework, and AI coding agents such as Claude Code, Codex CLI and Gemini CLI that need to verify their own output before it ships.
Kane CLI watches the rendered screen rather than waiting on network idle or DOM ready, so it works on canvas apps, shadow DOM and pages that stream their UI, where traditional waits break. When a button gets a new CSS class or a label is reworded, it re-resolves the target without a code change. Every element match carries a confidence score, so a low-confidence match is rejected instead of clicking the wrong thing and passing quietly.
Every run produces a ShareLink, a public URL valid for seven days with the run summary, screenshots and full step trace, ready to drop into a pull request or bug report.
Run it three ways: an interactive terminal UI, a scriptable CLI inside CI/CD, or agent mode that streams structured NDJSON events. Import existing Playwright or Selenium scripts, or export any run back to native Python Playwright code.
Install with npm and start free.
It is built for developers who want to check a change before raising a pull request, QA teams who want browser coverage without maintaining a framework, and AI coding agents such as Claude Code, Codex CLI and Gemini CLI that need to verify their own output before it ships.
Kane CLI watches the rendered screen rather than waiting on network idle or DOM ready, so it works on canvas apps, shadow DOM and pages that stream their UI, where traditional waits break. When a button gets a new CSS class or a label is reworded, it re-resolves the target without a code change. Every element match carries a confidence score, so a low-confidence match is rejected instead of clicking the wrong thing and passing quietly.
Every run produces a ShareLink, a public URL valid for seven days with the run summary, screenshots and full step trace, ready to drop into a pull request or bug report.
Run it three ways: an interactive terminal UI, a scriptable CLI inside CI/CD, or agent mode that streams structured NDJSON events. Import existing Playwright or Selenium scripts, or export any run back to native Python Playwright code.
Install with npm and start free.