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
FuseBase
Congrats, team! How does Kane decide whether a natural-language test has actually passed when the page behavior is slightly different than expected?
TestMu AI
@kate_ramakaieva Thanks Kate! A pass is never vibes. It is granted only when the expected state is verified through explicit evidence: DOM state, URL changes, network responses, screenshots. If the page differs cosmetically (moved button, new label), Autoheal re-resolves the intent and continues. If the expected outcome itself is not met, or the match is genuinely ambiguous, it fails loudly. Real bugs still fail.
TestMu AI
@kate_ramakaieva
Cosmetic drift like a reworded label or a new CSS class gets re-resolved, since intent is anchored to what the user sees. Anything ambiguous is rejected rather than guessed at.
The pass itself needs evidence on the live page: DOM state, URL change, network response, or your own assertion. If there is no evidence, then there would be no green, and the .evidence pack itself will shows you the exact step where it broke.
Kane CLI is the future of agentic verification systems.
TestMu AI
@chaitgpt1 Thank you Chaitanya 🙌 Agentic verification is exactly the category we are building. The agent builds it, Kane CLI proves it works.
Testing across browsers and real devices already creates plenty of complexity. Adding agentic reasoning could make that easier to manage.
TestMu AI
@bastienrenard That's exactly the bet 🙌 The matrix is rarely the hard part, maintaining it is. One DOM change and you're patching selectors across every target.
Agentic reasoning keeps the objective stable and lets the agent resolve the path each run. Which is also why every Kane CLI run ships with evidence, not just a green tick ✅
TestMu AI
Hi, I'm Mayank Bhola, Co-Founder and Head of Products @TestMu AI
It started with a pattern that would not go away. Agent ships the code. PR merges. Tests pass. Three days later someone opens the app and the button does not work.
Everyone kept asking how to make agents write better code. Wrong question. Agents could already open browsers and check their work. The output was the problem. Different result every run. No verdict you could trust. Loop it enough times to be sure and you have burned real money for a maybe.
So we set one non-negotiable bar: determinism
Same flow, same result, every time. A pass is granted only when the expected state is verified through explicit evidence:
DOM state and URL changes
Network responses
Screenshots and console logs
AC coverage behind every verdict
And as of today, that bar holds on mobile too
v0.8.1 brings the same verification contract to Android emulators and iOS simulators. Same evidence, same replay, same verdict you can defend.
This is the layer I wished existed every time a merged PR broke in production. Now it does, on web and mobile.
Break it. Question it. I am here all day.
Learn more: testmuai.com/kane-cli-ph
Does the product integrate with other systems?
TestMu AI
@dan_rotaru Yes! CI/CD out of the box (GitHub Actions, GitLab, Jenkins) via exit codes, coding agents like Claude Code and Cursor via structured NDJSON output, and two-way Playwright/Selenium: import your existing scripts or export any run to native Playwright code.
TestMu AI
@dan_rotaru Yes it does.
CLI interface itself opens up a lot of possibilities and it can be wired up with a large number of dev tools out of the box.
We also release more integrations every other day so if you have any specific tool in mind, let us know and we will be happy to release integrations.
Voquill
This is super useful, especially with how much code is being written by AI now.
TestMu AI
@henry_habib Thanks Henry! That is exactly the gap. AI made writing code fast, verifying it stayed manual. Kane CLI closes that loop.
Could developers trigger Kane directly from CI/CD and block deployments automatically when critical browser workflows fail?
TestMu AI
@himani_sah1 KaneCLI can directly run from CI/CD pipelines and allow you to create validation harness. If critical flows break, you changes can be blocked. More information available here: https://www.testmuai.com/docs/kane-cli-cicd
Lancepilot
TestMu AI
@priyankamandal Kane CLI does not fail silently or try to bypass these. It has a built-in Ask tool: when a run hits an OTP screen, a CAPTCHA, a payment confirmation, or any ambiguous state, it pauses, asks the human for input, and resumes exactly where it left off once answered.
TestMu AI
@lt_prakhar Yes! The Starter plan is free with 100 credits, no framework setup. One install command and your first run lands in minutes 🚀
TestMu AI
@lt_prakhar It's free to signup and use. It comes with free credits for you to get started with.
But would be happy to open up broader access if required for POC :)
ProductBridge
The build-fast, trust-slow gap is very real. I like how Kane CLI focuses not just on running tests, but on producing evidence teams can actually trust and share.
How do you see Kane CLI fitting into an AI coding workflow,should testing happen automatically after every code change?
TestMu AI
@hareesh_vemasani Yes, exactly, verification after every change, not as a phase at the end. Your coding agent calls Kane CLI after writing code and fixes its own bugs before you look, and in CI it acts as a deploy gate that blocks merges when critical flows fail. When building is this fast, verification has to be just as continuous.
TestMu AI
@hareesh_vemasani Ideally yes, AI and AI powered infra has changed the game altoghere. Now we have the rights tools to ensure that every code change carries with it necessary evidence showcasing that it does what it was supposed to do. And not only at unit level but also at integration and end-to-end level. Kane CLI fills that gap.
It can also sit in CI pipes and can be part of every git pull request.