Kane CLI - Natural language browser & mobile app tests from terminal
by•
An agentic quality verifier for developers and AI coding agents. Describe a test in natural language, and Kane CLI runs it in a real Chrome browser and returns pass or fail with shareable proof. No selectors to write. Local-first, free to start.
Replies
TestMu AI
Hi, I'm Jay Singh, Co-Founder of TestMu AI.
Every technology wave has the same shape. Capability jumps first. Trust catches up later. We saw it with cloud, and we are watching it again with AI.
The gap we kept seeing
⚡ AI writes the code in minutes now
🖱️ But someone still has to open the app and confirm it actually works
⏳ The building got fast. The assurance never did
That gap is the defining problem of this era of software, and it is why we built Kane CLI.
How it closes the loop
📥 Ingests your source or PRD
🧩 Designs the use cases and test scenarios
🌐 Runs them in a real Chrome browser, or on Android emulators and iOS simulators
📸 Returns an evidence pack, AC coverage, and a verdict
Not just "did it run". Did it work, and can you prove it. ✅
We believe the next generation of software teams will not be measured by how fast they build. Everyone will build fast. They will be measured by how fast they can trust what they built. 🚀
Would love to hear how your team is dealing with this gap.
Learn more: https://www.testmuai.com/kane-cli-ph/
Nas.com
How much context can Kane understand from a PRD before generating the actual browser scenarios?
TestMu AI
@nuseir_yassin1 Kane doesn't read the PRD as one giant prompt. Durable context - domain terms, roles, business rules - goes into a project context file that applies to every run. Each scenario is then one objective, with multiple acceptance criteria. The constraint isn't context window, it's decomposition.
The limit isn't how long your PRD is, it's how well it decomposes into self-contained scenarios.
What has been the most surprising use case you've seen from early users?
Congrats @jay_0687 & team!
TestMu AI
@hamza_afzal_butt Ooh fun question thanks Hamza 🙏
Most fun one: a user runs Kane CLI with their agent to order coffee on a schedule ☕😂 we built a quality verifier, he built a barista.
Most clever one: teams dropping Kane CLI into PR CI as a critical path check. Every PR gets asked "does the money flow still work?" in a real browser before it merges. Synthetic monitoring, shifted left into the review loop 🫡
how is this different from Claude in Chrome?
TestMu AI
@naima_dar Kane-CLI does everything Claude in Chrome does: drives a real browser from natural language, navigates, fills, clicks, verifies. The difference is what it leaves behind.
Claude in Chrome reasons its way through the task every time you ask. Kane does that once, then hands you a deterministic test case: exported code, structured NDJSON, per-step logs and screenshots, all persisted to Test Manager. The next thousand runs replay that artifact instead of re-deriving it. Same result every time, in CI, with a real exit code.
That's the future-proofing. Agentic exploration is how you author the test; determinism is what makes it a regression suite you can trust six months from now.
TestMu AI
Hi, I'm Mudit Singh, Co-Founder of TestMu AI.
The thing I love most about Kane CLI: the whole loop works for everyone.
👨💻 Developers feed it a PRD and get scenarios, runs, and a verdict before the PR
✍️ Business teams can describe flows in natural language and automate the boring browser stuff
🤖 AI agents run it to verify their own work and read the verdict straight from the output
📱 Mobile teams, starting today: point the same commands at Android emulators and iOS simulators
Every run leaves an evidence pack. Every verdict is backed by AC coverage. Nothing rests on "trust me, I checked".
Getting started is the easy part
The Starter plan is free. One install command, npm, brew, or curl, and your first verdict lands in minutes.
Try it, break it, tell us what is missing. Your feedback this week directly shapes the roadmap.
Learn more: testmuai.com/kane-cli-ph
TestMu AI
Hi, I'm Shantanu Wali, Head of AI Products at TestMu AI.
Most tools give you a test runner. We built an assurance loop:
📥 Ingest → your source, your PRD
🧩 Design → use cases, acceptance criteria, test scenarios
▶️ Run → test.md, executed in a real browser or on a mobile device
📸 Evidence → evidence pack for every run
📊 Coverage → how many ACs are covered
✅ Verdict → shipping confidence, backed by proof
Decisions we fought hard for
💻 Local first, running on your own Chrome
💬 Natural language objectives instead of selectors
🔧 Autoheal, so cosmetic UI changes never break a flow
🤖 Agent-native output, so Claude Code or Cursor can call Kane CLI and read the verdict on their own
And starting today, mobile 📱
v0.8.1 makes mobile a first-class platform. Same setup, same sessions, same evidence model, now on Android emulators and iOS simulators. Run devices list, pick a target, and the AI reasons about mobile screens the way it reasons about web pages.
For the skeptics
Hacker Mode exports any flow to native
Playwright code. Never locked in. 🔓
Ask me anything technical, I will be here all day.
Learn more: testmuai.com/kane-cli-ph
I really love KaneAI’s conversational approach to generating critical test scenarios from the provided documentation. It frees up my local machine’s resources while helping me explore and work with different aspects of the platform more efficiently. Overall, KaneAI has made the testing process smoother, faster, and more collaborative.
TestMu AI
@anmol_soin1 Thank you Anmol, love hearing this 🙌 KaneAI keeps doing exactly that, conversational scenario generation from your docs, and it is going strong. The fun part: Kane CLI shares the same engine, brought to the terminal. Same intelligence, now living where developers and coding agents work, with an evidence pack and verdict on every run. Two surfaces, one engine, and the workflow you described gets faster on both.
TestMu AI
@anmol_soin1 That really motivating to hear :)
Glad that our platform could make your testing process easier.
Hi everyone, I'm one of the engineers building Kane CLI. Good to finally have this out in the world.
The part I'm personally most proud of: we built it agent-native from day one. I've spent 2+ years shipping frontend at TestMu AI, and if that taught me anything, it's that writing UI was never the slow part, proving it still works after every change is. AI has made the writing nearly instant, which makes verification the real bottleneck. So we designed for it: every command can emit machine readable output, meaning it's not just humans in a terminal, your coding agent (Claude Code, Cursor, whatever you use) can invoke kane-cli, stream the run, read the verdict, and act on it inside its own loop. AI writes the code, Kane CLI verifies it in a real browser (or on mobile), and the agent gets machine-readable proof back. Closing that loop is what makes autonomous dev workflows actually trustworthy.
The other thing we obsessed over: a pass has to actually mean something. Every verdict ships with an evidence pack; screenshots, console logs, network responses; so a green check is never "the model felt good about it." You can see exactly what was verified, and when something fails you're not guessing.
It's local-first and free to start. Happy to answer any questions in the comments.
the ask-tool pause on OTP/CAPTCHA instead of guessing is the right call. since the exported case replays deterministically, what happens when the site's UI shifts a little, a moved button or new copy, does it re-derive from the original NL description or does someone have to re-record?
TestMu AI
@sabber_ahamed That is handled a level above element healing. Run with --retry and Kane CLI auto-detects when the flow itself has changed, a new step in the journey, a reordered screen, and adapts the run to the live flow instead of dying on the old script. Checks heal on replay as well, so the test.md keeps working without anyone re-authoring it.
TestMu AI
@sabber_ahamed re-derive from the original NL description. The natural language test becomes the source of truth.
TestMu AI
Big day for us at TestMu AI
I have watched Kane CLI come together from the inside, and the part that still gets me is watching it in action. You type what you want verified in natural language, a real Chrome window opens, it works through the flow, and comes back with an evidence pack and a verdict. No selectors written. No framework set up.
And the timing of this launch could not be better. As of today it does the same thing on Android emulators and iOS simulators. Same commands, same evidence, now on mobile.
The team has poured months into getting the details right, determinism, Autoheal, agent-native output, and it shows.
Proud of this one. If you ship with AI agents, or you are just tired of clicking through your app before every release, give it a run. First verdict lands in minutes.
Learn more: https://www.testmuai.com/kane-cli-ph/