TestMu AI Launches Kane CLI — Terminal-First Browser Automation Built for AI Agents and Developers (April 28, 2026)
TestMu AI (formerly LambdaTest) on April 28, 2026 launched Kane CLI, a free terminal-first browser automation tool with three modes — Interactive TUI, Headless CLI, and a dedicated Agent Mode that emits structured NDJSON for Claude Code, Codex CLI, and Gemini CLI. It is the first major browser-test runner explicitly built for both humans and coding agents.
TestMu AI — the agentic quality-engineering platform formerly known as LambdaTest — on launched Kane CLI, a free terminal-first browser automation tool built simultaneously for human developers and AI coding agents, available immediately via npm and Homebrew.
What Happened
Announced on PR Newswire on April 28, 2026 and detailed on the official Kane CLI page, Kane is positioned as the first major browser-test runner with a dedicated Agent Mode — a flag (--agent --headless) that emits structured NDJSON output explicitly designed for Claude Code, OpenAI Codex CLI, and Google Gemini CLI to consume. The launch closes a gap that has plagued AI-assisted development since coding agents went mainstream: agents could write Playwright or Selenium code, but had no reliable way to verify that the resulting flows actually pass in a real browser.
Kane CLI ships in three operating modes from the same binary. The Interactive TUI launches with no arguments and opens a full terminal UI for exploring tests against a live Chromium session, the Headless CLI runs scriptable display-free flows in CI pipelines, and the Agent Mode pairs --agent --headless with NDJSON output so an AI agent can read step-level results and decide what to do next. Local runs are free; cloud runs against the TestMu AI grid are billed against an existing TestMu AI plan.
Key Details
- Three operating modes — Interactive TUI, Headless CLI, and Agent Mode — all in one binary, switched by flags so humans and agents consume the same tool.
- Resilient runs up to 50 steps per flow — Kane operates on intent rather than CSS selectors and adapts through transient failures, retrying until the user journey is verified or definitively fails.
- Vision-based dynamic waiting — instead of network-idle heuristics, Kane looks at the rendered page to detect loaders, animations, and skeleton states before acting, which fixes a long-running source of Playwright and Cypress flake.
- Playwright export — a single command converts a passing Kane flow into native Playwright test code, so teams can use Kane for authoring and Playwright for long-term maintenance.
- Native CI integrations — ships GitHub Actions, GitLab CI, Jenkins, and Bitbucket Pipelines support out of the box with display-free headless runs.
- Free to start — install via
npm install -g @testmuai/kane-cliorbrew install LambdaTest/kane/kane-cli; cloud-grid runs billed against TestMu AI plans. - Three months of bonus credits for teams activating a paid TestMu AI plan during the launch window.
What Developers Are Saying
Initial reaction from quality-engineering channels has focused on Agent Mode as the differentiator: most AI coding agents in 2026 can write a Playwright test, but verifying it without bouncing through a heavyweight orchestration layer has been the unsolved problem. The structured NDJSON output gives agents a deterministic way to read step-by-step results, which lines up with how Anthropic's Claude Code, OpenAI's Codex CLI, and Google's Gemini CLI already consume tool output. The recurring skepticism is the same one that has greeted every "intent-based" testing tool: vision-based waits and 50-step resilient flows reduce flake on simple journeys, but complex multi-tab e-commerce checkouts and authenticated SaaS flows remain the hard cases that have broken every prior generation of low-code testing platforms.
What This Means for Developers
For teams already running CI on GitHub Actions or GitLab, Kane CLI is a low-risk add-on: install it via npm or brew, run a flow, and if you like the output, use the Playwright export to commit a stable test suite to your repo. For teams building agent-driven development workflows around Claude Code, Codex CLI, or Gemini CLI, Kane's Agent Mode is the first publicly available browser-test runner with a documented NDJSON contract aimed specifically at agent consumption — meaning agents can self-verify UI changes without a custom test harness. The free local-run tier means there is no procurement step to get started.
What's Next
TestMu AI says Kane CLI source is browsable on GitHub, with public issue tracking and a roadmap for community contributions. The cloud-grid integration is gated behind existing TestMu AI plans, which means teams already paying for LambdaTest's grid get Kane runs on the same hardware. Expect a steady cadence of integrations — the agent-mode story will only matter if the NDJSON contract stays stable across releases as Claude Code, Codex CLI, and Gemini CLI iterate on their tool-use formats.
Sources
- TestMu AI — Kane CLI official page — primary source with feature documentation and install instructions.
- PR Newswire — Launch announcement (April 28, 2026)
- AIThority — Independent coverage with technical detail
- IT Brief — Kane CLI for terminal browser tests
- StartupHub.ai — Kane CLI bridges AI code to verified browser actions
- TestMu AI — Rebrand from LambdaTest (January 12, 2026) — background on the parent company.
Stay up to date with Doolpa
Subscribe to Newsletter →