Kane CLI By TestMu AI

Kane CLI is a terminal-native browser testing tool built for developers, QA engineers, and AI coding agents. Describe a browser flow in plain English and Kane CLI executes it in a real Chrome browser, verifies every step, and returns a clean pass or fail. No selectors to write, no test scripts to maintain, no framework to learn from scratch.

The Problem It Solves

AI coding agents like Claude Code, Codex CLI, and Gemini CLI can write features, fix bugs, and refactor code at impressive speed. What they cannot do is open a real browser, navigate to the application, and confirm that what they built actually works end to end. They can run unit tests. They cannot verify a multi-step browser flow with real DOM, real network requests, and real user interactions. Kane CLI fills that gap.

For human developers, the problem is different but equally real. Verifying a feature before a PR means either manually clicking through the flow in a browser or waiting for a full CI pipeline to run. Neither is fast enough to be a natural part of a development loop. Kane CLI gives developers a third option: run the verification from the terminal in under two minutes, before the PR is even opened.

Three Execution Modes

Kane CLI ships with three execution modes designed to fit different workflows.

Interactive TUI is the default mode. The browser stays open between runs, so session state carries over. Authenticate once and subsequent runs start from that authenticated state. This is the mode for development and exploratory testing.

Headless CLI runs a flow with no visible browser. It produces a deterministic exit code — 0 for pass, non-zero for fail — and output goes to stdout. This is the mode for shell scripts and CI/CD pipelines.

Agent Mode outputs structured NDJSON, one JSON object per line, each representing a step in the execution. This output is designed to be consumed by AI coding agents and automated systems that need to parse and act on results programmatically.

Built for AI Coding Agents

Setting up an AI coding agent to use Kane CLI takes one command:

npx @testmuai/kane-cli-skill

This installs a SKILL.md and AGENTS.md file that tells Claude Code, Codex CLI, or Gemini CLI exactly how to use Kane CLI, the available commands, the expected inputs, the output schema, and example patterns. The integration is one command away from being operational.

Key Capabilities

The automation engine is vision-based rather than selector-based. It identifies elements by what they look like on screen, not by CSS classes or DOM position. This makes it resilient to the UI changes that break traditional selector-dependent automation. Built-in autoheal and smart waiting detect loaders and animations before acting.

Test.md is Kane CLI's agent-native test format. Any session can be saved as a replayable markdown file that is human-readable, version-controllable, committable to a repository, and executable by AI agents. This is the test framework designed for the era of AI-assisted development.

Native Playwright export lets any Kane CLI flow be converted to an idiomatic Playwright script. Existing Playwright and Selenium scripts can also be imported into Kane CLI.

Every run generates a shareable evidence link with the full session recording, step-by-step logs, and complete network and console capture, synced automatically to the KaneAI Test Manager dashboard.

Pricing

Free tier: 200 credits per month, no card required. Paid plans from 19 dollars per month. Enterprise available with SSO, IP whitelisting, and compliance support.

Available on macOS, Linux, and Windows. Install with:

npm install -g @testmuai/kane-cli