Kane CLI by TestMu AI

@mikenevermiss The loop is only as honest as its verify step. Claude Code verifies its own code. Tests pass, types check, "passed," but it never sees the UI render, so a broken button still ships green. Kane CLI is the verify step in real Chrome. pass or fail on what users actually see.

this is f*cking dangerous someone just open sourced the entire "LOOP ENGINEERING" framework for free the guy who built Claude Code at Anthropic said it himself: "I don't prompt Claude anymore. I have loops running that prompt Claude. My job is to write loops. stop prompting your agent. build the thing that prompts it for you. discover → plan → execute → verify → repeat you used to be inside that loop loop engineering is you stepping out of it entirely the repo includes: - 6 production ready patterns (daily triage, CI sweeper, PR babysitter) - clone and run starters for Claude Code, Codex, and Grok - a loop readiness CLI that scores your codebase - SKILL. md templates, STATE.md spine, full safety docs use this one command to start(copy/paste): npx @cobusgreyling/loop-init • --pattern daily-triage -tool claude-code bookmark before this blows up repo link : github.com/cobusgreyling/…

Now you can watch it think. Give Kane CLI the objective, and it reads the screen, plans the steps, and verifies the result live. A pass it earns in front of you, not just claims at the end. Faster runs, and full visibility into every move. Register now 🔗 testmuai.com/register/?utm_… #AI #BrowserAutomation #KaneCLI

Kane CLI just shifted gears 🏎️ It's now ~3X faster, with the resilience to match.

Speed is part of correctness. A check fast enough to run before every PR catches things while they are still cheap to fix. 🏃‍♂️✅ #KaneCLI

$ 𝙺𝚊𝚗𝚎 𝙲𝙻𝙸 --𝚝𝚒𝚙 𝟷𝟺💡 Match the window size to your real users. 𝚔𝚊𝚗𝚎-𝚌𝚕𝚒 𝚌𝚘𝚗𝚏𝚒𝚐 𝚜𝚎𝚝-𝚠𝚒𝚗𝚍𝚘𝚠 𝟷𝟹𝟼𝟼𝚡𝟽𝟼𝟾 Test what users actually see. Catch layout bugs that only show up at their screen size. #QA #KaneCLI

$ 𝙺𝚊𝚗𝚎 𝙲𝙻𝙸 --𝚝𝚒𝚙 𝟷𝟹💡 Skip code validation when you just want a draft. 𝚔𝚊𝚗𝚎-𝚌𝚕𝚒 𝚛𝚞𝚗 "．．．" --𝚌𝚘𝚍𝚎-𝚎𝚡𝚙𝚘𝚛𝚝 --𝚜𝚔𝚒𝚙-𝚌𝚘𝚍𝚎-𝚟𝚊𝚕𝚒𝚍𝚊𝚝𝚒𝚘𝚗 Faster output when you need a starting point, not a polished file. #KaneCLI

Claude vibe codes against text: tests, types, linter, compiler. All read the code, none open a browser. So "passed" just means it compiled, not that the button works. CI inherits the same blind spot. Kane CLI fixes the boring way: run the real flow in a real browser on every push, pass or fail.

the worst part of using any software from frontier labs is just how broken it is, no attention to detail, just vibe coded slop ud think they could have people do some QA before it hits prod or have some edge cases u can come up with

@vicky_grok Evaluate is where these loops break down. Kane CLI ties it to a browser run you describe once, with a clear pass or fail, so you get the same result every time, even as agents scale.

kane-cli 0.4.8 ✨ 🌟 WebSocket capture, now surfaced end-to-end 🧪 Smarter AI-generated test steps 💫 Version check that actually works 🪟 expose WebSocket capture in TUI Config > Run Tests, replayed. github.com/LambdaTest/kan…

$ 𝙺𝚊𝚗𝚎 𝙲𝙻𝙸 --𝚝𝚒𝚙 𝟷𝟸💡 Give it a hint about where to look. kane-cli run "assert the price in the Pro card, under the plan name, shows '$19'" Busy pages have many prices. A location hint stops the agent from grabbing the wrong one. Right answer on the first try. #KaneCLI

Do it agentic way: testmuai.com/kane-cli/?utm_…

"Register me for TestMu Conf 2026 and download my attendee card." The most agentic way to register for TestMu Conf 2026 😎 One line. Kane CLI filled the form, hit register, and pulled the card in the same session. No clicks. No re-login. Register now for TestMu Conf 2026 🔗 testmuai.com/testmuconf-202… #BrowserAutomation #KaneCLI

$ 𝙺𝚊𝚗𝚎 𝙲𝙻𝙸 --𝚝𝚒𝚙 𝟷𝟷💡 Pick the right mode. Action does the task. Testing grades it. kane-cli run "fill the signup form" --mode action Use action when you just want something done. Use testing when you want a strict pass-or-fail gate. Testing turns every check into a hard assertion, so do not reach for it when you only meant to click around. #KaneCLI

Kane CLI gives your agent eyes. 👁️ One command, agent mode: ⌨️ kane-cli run "go to localhost:3000, click Buy Pro, assert the total due today shows $19.00" --agent Kane CLI spins up real Chrome, clicks through to checkout, and makes the call. Advertised $19. Charged $29. Failed, with a report you can share. Gemini CLI takes that report, opens page.js and checkout/page.js, finds the hardcoded $29, and patches it. A second Kane CLI run checks the fix. ✅ Pass. A hardcoded value is the easy case. A price pulled from an API. A coupon that applies on render. An overlay covering the total. None of that lives in the code. The browser is the only ground truth. Reasoning plus verification. The loop closes itself. Write, test, fix, verify, all in one terminal. Explore now 🔗 testmuai.com/kane-cli/?utm_… #GeminiCLI #KaneCLI

Your agent ships a feature in 4 minutes. You verify it in 30. Kane CLI makes API calls first-class test steps. One run hits your backend and your frontend, threading the API response straight into the browser checks. No glue code. No separate API and UI suites. One natural language prompt: Kane CLI called the GitHub API, opened a real browser, read the latest release tag on the page, and asserted it matched the API response. New in 0.4.6: → External APIs as first-class test steps → Thread API responses into later steps and assertions by variable → Child flows inherit API context across nesting levels → Failed calls surface as failures, nothing dropped silently API testing and browser automation. Natural language. From your terminal. Start free 🔗 testmuai.com/kane-cli/?utm_… #KaneCLI #TestMuAI

Screenshots help. A replayable run helps more, because the reviewer can watch the whole thing instead of imagining it. 🔁📼

$ 𝙺𝚊𝚗𝚎 𝙲𝙻𝙸 --𝚝𝚒𝚙 𝟷𝟶💡 Match what is on the screen, not what is in your head. 𝚔𝚊𝚗𝚎-𝚌𝚕𝚒 𝚛𝚞𝚗 "𝚐𝚘 𝚝𝚘 𝚑𝚝𝚝𝚙𝚜://𝚖𝚢-𝚝𝚎𝚜𝚝𝚒𝚗𝚐-𝚛𝚎𝚙𝚘-𝚖𝚊𝚒𝚗．𝚟𝚎𝚛𝚌𝚎𝚕．𝚊𝚙𝚙, 𝚊𝚜𝚜𝚎𝚛𝚝 𝚝𝚑𝚎 𝙿𝚛𝚘 𝚙𝚛𝚒𝚌𝚎 𝚜𝚑𝚘𝚠𝚜 '$𝟷𝟿 / 𝚖𝚘'" Kane CLI reads the page like a person. Write "$19/mo" when the page says "$19 / mo" and the check fails on wording alone. Copy the exact text. No false alarms. #QA #KaneCLI

kane-cli 0.4.7 ⚔️ 🪟 A live run view that shows what's happening ✨ More accurate assertions and text extraction 🌟 Cleaner, more accurate step labels 🐛 Reliability and display fixes Sharper than last time. github.com/LambdaTest/kan…

Cursor writes the code. Kane CLI proves it works. ✅ Both from the same terminal, no context switching. Type what you want checked in natural language, and Kane CLI drives a real Chrome browser to verify the flow end to end. 🌐 Take the hardest case: an OTP login. Most tools choke the moment a real one-time code is involved. 🔐 Kane CLI runs the flow, and when it hits the code, it just asks you for it, then carries on to verify and finish. What you get back is not a one-off. Every run lands in Test Manager as a clean, replayable test case you can rerun any time. 🔁 Less clicking through flows by hand. More shipping with proof. 🚀 Run your first test free 🔗 testmuai.com/kane-cli/?utm_… #Cursor #KaneCLI #TestMuAI

Automation and testing can live in one place, and you are not locked in either. Prototype the flow in plain English, then export it to native Playwright when you want it inside your existing suite. 🧰📤 #AutomationTesting #KaneCLI