Kane CLI by TestMu AI

End-to-end flows on your local browser, powered by natural language. Instant validation. Deep bug discovery. Automation that ships with confidence.

[Kane CLI by TestMu AI @kane_cli]

kane-cli 0.4.6 🚀
🧪 API steps inside test flows
✨ Capture and observability
🌟 Reliability and failure surfacing
🔧 Tool access for nested child flows
APIs, executed!
https://t.co/KNFr0WhuKZ

[Kane CLI by TestMu AI @kane_cli]

A login wall does not have to stop automation.
Treat the OTP as a quick pause, and the flow keeps going.💻⏸️

#KaneCLI

[Kane CLI by TestMu AI @kane_cli]

$ 𝙺𝚊𝚗𝚎 𝙲𝙻𝙸 --𝚝𝚒𝚙 𝟶𝟾💡

Keep credentials out of your objectives. Pass them as variables and reference with {{key}}.

kane-cli run "log in with {{email}} and {{password}}" --variables-file creds.json

Mark secrets with "secret": true so they never get committed. Same flow runs on dev, staging, and prod, only the values change.

#KaneCLI

[Kane CLI by TestMu AI @kane_cli]

Before a PR goes up, the feature should already be verified in a real browser, locally, before anyone reviews it.

Here is what that catches.

A user clicks "New project", types a name, hits Create. The modal closes. Looks done. Kane CLI watches the sidebar instead, and the new project never appears. Run failed.

The replay points straight at the create-project flow. 🔍

The project was built in memory, then dropped before it ever rendered. The code ran. The modal closed. Nothing threw an error. The UI was still wrong. That is the gap a passing unit test never sees.

Fix it. Run the check again. Green. ✅

Now the PR goes up, already verified in a real browser. Not caught in review. Not in staging. In the loop, before.

One objective in natural language. Pass or fail out.

Try it on your own pre-PR loop 👉 https://t.co/C8OZYW9e1K

#AIAgent #KaneCLI

[Kane CLI by TestMu AI @kane_cli]

https://t.co/5oLdEtTDuh

[Kane CLI by TestMu AI @kane_cli]

@nichochar Tokens go brrrr because the agent is paying (in your tokens) to click around a browser. that's the bottleneck. Kane CLI offloads the browser proving to its own engine, agent just reads a pass/fail. token bill stops climbing.

[Kane CLI by TestMu AI @kane_cli]

@Hartdrawss Wild find. "Production ready" is just a claim until something runs the flow end to end. That's where we exactly fit in.

[Kane CLI by TestMu AI @kane_cli]

$ 𝙺𝚊𝚗𝚎 𝙲𝙻𝙸 --𝚝𝚒𝚙 𝟶𝟽💡

Headless CLI mode, you pass the objective straight into the command: kane-cli run "...". One line in, a real browser verdict out.

🔁 Repeatable by design. Arrow up, hit enter, run the same flow again in a second. Same command, same objective, no drift between runs.

#AIAgents #KaneCLI

[Kane CLI by TestMu AI @kane_cli]

Use Kane CLI in Claude Managed Agents.

Kane CLI already runs in Claude Code. Now add it in Claude Managed Agents too.

The same agent that builds a feature can open a real browser and verify it. Pass or fail, with a shareable link.

Set it up now:
✅Create an environment and install kane-cli
✅Store creds in a Claude credential vault. The agent never sees the values
✅Define the agent in one line: run browser tests, report pass or fail
✅Give it a URL and a flow in natural language

Try it 🔗 https://t.co/xCmvimlgNx

[Kane CLI by TestMu AI @kane_cli]

Verification belongs at the moment of building as well, not only in a QA phase a few days later.

Close the gap where you can. ⚡ 🔁

[Kane CLI by TestMu AI @kane_cli]

$ Kane CLI --tip 06💡

Set --max-steps to match the flow. Simple check, low cap. Complex journey, higher cap.

Size it just above a clean run: kane-cli run "..." --max-steps 10

Benefit: a wandering agent stops early, real flows still finish, and you stop burning credits on runaway runs.

#BrowserAutomation #AI #AITesting #KaneCLI

[Kane CLI by TestMu AI @kane_cli]

kane-cli 0.4.5 ⚔️
✨ Smarter retries and timeouts
🌟 More reliable variable and URL handling
💫 Cleaner run lifecycle
🔁 bounded retry + configurable timeout for Chrome CDP…
Sharper than last time.
https://t.co/9rCIWiTueE

[Kane CLI by TestMu AI @kane_cli]

Agent mode speaks NDJSON. This is why it works. 📡

We gave Claude Code one natural language objective: open a live /cart page and check for console warnings.

It called Kane CLI in agent mode, ran it in real Chrome, and streamed back NDJSON. One JSON object per line. No screenshot, no log dump.

The run_end line carried everything: status, captured warnings, extracted values, duration, and a dashboard link. Result: PASSED. ✅

An agent cannot act on a picture of a page. It can act on structured NDJSON.
The agent builds it. Kane CLI proves it works.

Try Kane CLI. Free to Start 🔗 https://t.co/4JXFqHhvpc

#BrowserAutomation #ClaudeCode #QA #AIAgents #AI #KaneCLI #TestMuAI

[Kane CLI by TestMu AI @kane_cli]

A browser agent that completes a task but cannot say pass or fail is a party trick.🎩

Verification has a verdict. Tasks do not.⚖️

#KaneCLI

[Kane CLI by TestMu AI @kane_cli]

$ 𝙺𝚊𝚗𝚎 𝙲𝙻𝙸 --𝚝𝚒𝚙 𝟶𝟻💡

Run Kane CLI's Interactive TUI mode where the browser stays open between runs, and your state carries over.

If you login once, the chain tests after test in the same session. No re-authenticating. No re-navigating to where you were. Cookies, session, and current page all persist from one run to the next.

#KaneCLI

[Kane CLI by TestMu AI @kane_cli]

kane-cli 0.4.4 🚀
✨ One place to set your start URL
🌟 Clearer errors when a URL is missing
💫 Fewer surprises in long sessions
🎉 Linux arm is now supported
Sharper than last time.
https://t.co/5XKwDQ9b5A

[Kane CLI by TestMu AI @kane_cli]

kane-cli 0.4.3 🛡️
🔍 New browser-automation tools
🌟 Isolated clipboard for the agent
🔐 Set Auth header via Set cookies
✨ Contextual hints in the footer
Tests that survive.
https://t.co/FtDFkmbgMf

[Shravan Mahajan @shvnmahajan]

https://t.co/7lzSqiDKAD

[Kane CLI by TestMu AI @kane_cli]

One objective. Three ways to run it.

Same natural language objective, all three Kane CLI modes:

🖥️ TUI Mode: type it, watch the step tree fill in. State persists, so the next run picks up where you left off.

⚡ CLI: kane-cli run "..." executes end to end, records the session, saves to KaneAI.

🤖 Agent Mode: Claude Code calls --agent --headless and reads structured step events directly.

Same binary. Same syntax. Your desk, a pipeline, or an agent's loop.

The agent builds it. Kane CLI proves it works.

Get started 🔗 https://t.co/FtW5mpoyD0

#BrowserAutomation #KaneCLI

[Kane CLI by TestMu AI @kane_cli]

Half your CI minutes go to fixing tests that broke because someone changed a color. 🎨

That is not a test suite. That is a tax.🧮