Kane CLI by TestMu AI

Agent Details

Kane CLI is a terminal native testing agent from TestMu AI, formerly LambdaTest. You describe what should happen in plain English, and Kane CLI opens a real Chrome browser, works through the flow, checks each step, and returns a pass or a fail you can act on. It was built for two audiences at once. Developers, QA engineers and anyone shipping fast get end to end browser coverage without writing selectors. AI coding agents such as Claude Code, Codex CLI, Cursor and Gemini CLI get a verification primitive they can call, so the code they generate can actually be proven to work in a browser instead of merely compiling. What makes runs hold up over time: Autoheal, smart waiting and dynamic querying keep tests passing when the UI shifts. Vision based waiting watches the screen for loaders and animations rather than guessing from network activity, which is what makes canvas, shadow DOM and heavily framework driven pages work. Kane CLI only takes actions a real person could take inside the browser viewport. It will not inject custom JavaScript to bypass a validation and force a green result. Test.md captures any session as replayable markdown, with imports, variables and replay. Tests live as readable text you can review in a pull request. Native Playwright export means you are never locked in. Existing Playwright and Selenium scripts can be converted in, and Kane CLI tests can be converted back out. Every run syncs to the KaneAI dashboard with step by step logs and a shareable evidence link, so proof is attached automatically. Three modes cover the workflows people actually have. An interactive TUI for development where the browser stays open between runs. A headless non interactive mode for shell scripts and CI. An agent mode that suppresses the TUI and emits structured NDJSON, one JSON object per line, for AI agents to parse. Install with npm, sign in, and run your first flow in a couple of minutes.

Key Features

• Natural language browser automation on your local Chrome
• Deterministic pass or fail on every run
• Autoheal, smart waiting and dynamic querying keep tests alive through UI changes
• Vision based waiting that handles loaders, canvas and shadow DOM
• Secrets and variables for parameterized flows
• Ask tool to unblock a run on OTPs and CAPTCHAs
• Custom profiles and reuse of existing authenticated sessions
• Shareable test evidence links with step by step logs