10,000 Kane CLI Credits. Yours. Free for a Month.
Claim NowWhite ArrowWhite Arrow

Kane CLI Testing Agent for DEVS

Kane CLI is an "agentic quality verifier" which executes software the way a user would, in a real browser and on mobile, and returns an evidence-backed pass or fail you can defend.

npm install -g @testmuai/kane-cli
Start free

For Claude Code, Cursor or Codex:

~/shop
https://ecommerce-playground.lambdatest.ioChrome
Playground store cart (1)
iPhone$146.00Add to cart
Checkout as guest
guest@shop.test
Continue
Review: guest@shop.test · total $146.00
$ kane-cli --tui  Welcome toCommand-line interface  The validation layer for AI coding agents.❯ Check out without signing in. Verify the email and the total. Mode: Run | ● authenticated  ✓ navigate  https://ecommerce-playground.lambda…  1.3s
  ✓ click     "Add to cart"                         2.2s
  ✓ fill      email = {{email}}                     2.9s
  ✓ click     "Continue"                            3.4s
  ✓ assert    the email you entered is shown        1.8s
  ✓ assert    the total matches the cart            1.6s
  PASSED  6/6 steps  in 41.6s
  evidence evidence.lambdatest.com/r/2f9a1
passedexit 0

For every coding agent you build with

  • Claude Code
  • Cursor
  • Codex
  • Gemini CLI
  • GitHub Actions
  • Any CI pipeline
  • Claude Code
  • Cursor
  • Codex
  • Gemini CLI
  • GitHub Actions
  • Any CI pipeline
A Gift from TestMu AI

10,000 Kane CLI Credits. Yours. Free for a Month.

Automate your QE workflow from your terminal

KANE-CLI — TERMINAL

$ npm install -g @testmuai/kane-cli

$ kane-cli run "verify checkout flow on staging"

✓ 12 tests generated

✓ 12 passed (41s)

// What is Kane CLI

A testing agent for you.
A co-agent for your coding agent.

$ kane-cli
ChromeAndroidiOS

Drive web and mobile in one sentence.

Browser automation and mobile automation in plain English, with no selectors.

See all five surfaces
4
replay · step 3Failed

Replay the crash, step by step.

Every step, screenshot and log is recorded, so you can walk back through any failure.

Learn more about evidence
REQ 1Proven
REQ 2Proven
REQ 3Still open

Prove it, requirement by requirement.

Link tests to your requirements. See which ones a real run has proven.

Learn more about coverage

THE ASSURANCE LOOP

From spec to proof.

One requirement, traced through all four: Users can reset a forgotten password.

1PlanASSURANCE
spec · password reset
AC-1AC-2AC-3
JiraConfluenceLineardoc

Reads your spec and lists what must be true. You approve it.

2DesignASSURANCE
reset link is emailed→ AC-1
new password signs in→ AC-2
old password is rejected→ AC-3

Writes one test per scenario, each linked to its requirement. You review them.

3ExecuteEVIDENCE

$ kane-cli testrun run

✓ reset link is emailed

✕ new password signs in

· old password: not run

Runs on your machine or the TestMu AI cloud, and seals the evidence.

4ReportEVIDENCE

coverage · 1 of 3 proven

AC-1 proven
AC-2 failed
AC-3○ still open

Shows what's proven, what failed, and what's still open.

When the spec changes, Kane CLI flags every test it affects.

See the assurance loop

// Why Kane CLI exists

Code ships faster than you can prove it works.

A quick click-through catches the obvious. It leaves nothing the next person can check.

A change is madeA quick click-throughNo result to hand over

The cost of unproven changes

testsmerged
testsnot merged

~50%

of agent PRs rejected despite green tests

Hand-written PR

Agent-written PR

5x

longer wait before an agent-written PR gets reviewed

3 in 10

main-branch builds fail

// How it works

Build. Validate. Prove. Repeat.

Everything your AI coding agent builds should go through this loop. It ends with a confidence score for the build.

> KANE CLIDefine: what must be true

// Inside a run

Your sentence, run on the real app.

Kane CLI plans the steps, performs each one in a real browser, checks the result, and returns a pass or fail with screenshots and logs.

YOU ASK

~/shop

$ kane-cli run

"Check out without signing in. Verify the email and the total."

Context

variables, secrets, saved login

Plan

steps and checkpoints

ON THE REAL APP

repeats for every step

ecommerce-playground.lambdatest.io
step 1step 2step 3step 4

Snapshot

DOM, accessibility tree, screenshot

Act

clicks and types like a user

Assert

DOM first, vision fallback

YOU GET

Passed· 2 of 2 checks

  • The email you entered is shown
  • The total matches the cart
  • Evidence packsteps, screenshots, logs, verdict
  • Test Manageruploads by default
  • PR commentwith the GitHub Action
  • Playwright codeon export

On replay, if the page changed, the broken step re-authors, and the change is recorded.

Learn more about evidence

// Why Kane CLI

Retire the selector graveyard.

Playwright makes you feel covered. Kane CLI proves you are.

1 · The test

Most of the Playwright tests didn't survive the night.

16 LIVE-SITE TESTS · 11 ROUNDS · TESTS PASSING, OF 16

Kane CLI 16 of 16 Playwright 2 of 16
0481216123456789101110 of 16 red by the next morning2 flipped back to green on their own

round

Pinned to day one

Every failure was a value frozen when the test was written: a story ID, a price, a headline.

Not repeatable

10 of 16 red by the next morning, 14 by round 7. Two flipped back to green on their own.

Costly to revive

12 broken tests took 12 AI repair sessions: $34.95.

Kane CLI: 16 of 16, every round.

It saves what each test means and heals the steps that move, so a changed price or headline does not break it.

2 · One change, two results

A button was renamed. One test broke, the other healed.

A coded test · sample

checkout.spec.ts

await page.getByRole('button', { name: 'Checkout' }).click();

await expect(page.getByTestId('review-email')).toHaveText(checkoutEmail);

✕ Timeout waiting for getByRole('button', { name: 'Checkout' })

// The button now says "Continue to payment". Someone edits the test.

The test breaks and waits for a person to edit it.

Kane CLI · sample

~/shop

$ kane-cli testmd run checkout_test.md

↻ Step 5 re-authored: "Checkout" is now "Continue to payment"

✓ Check: the order review shows the email you entered

✓ Passed · the re-authored step is recorded in the result

Heals, passes, and logs the change for review.

Re-authored steps are recorded for review. On a real failure, fix the app first.

3 · Included

Built into every run.

Resilient

Re-authors from the broken step

Persistent

Pause, crash, resume

Plain English

No framework to learn

Agent-native

Output any agent can read

Console and network checksSaved logins and profilesMasked secrets and variablesTests from a descriptionBug detectionHeadless or visible

If you already use Playwright, keep it. Any recorded flow exports as Playwright code.

PRISM benchmark · 2026-Q4

Ranked #1 for browser testing

50 real browser scenarios, scored on whether a pass means the flow worked.

See the leaderboard

Overall score

#1

76.7%

#1 of 7 tools

Reliability

88.7%

Same right answer, run after run

Perception

86.7%

Reads the page like a person does

Intent

77.9%

Does only what the test asked

// Surfaces

Web. API. Emulator. Simulator. Real device.

Terminal testing for all five surfaces, written in plain English.

Web

real Chrome

API

requests and responses

Android Emulator

native apps

iOS Simulator

native apps

Soon

Real device

physical phones

~/my-app · Android Emulator

$ kane-cli run "Log in and verify the home feed loads" --target emulator --app ./app.apk

the Mobile CLI takes the same plain-English objective as the Web CLI

same pass or fail, same evidence per step

For local runs

  • A Mac with Apple Silicon
  • Android Studio or Xcode
  • Your app build or an uploaded app ID

Without an Apple Silicon Mac, saved tests can run on the TestMu AI cloud, with the remote-execution plugin and a HyperExecute macOS plan.

Read the mobile setup guide

// Ecosystem

Keep your stack. Put proof at the center.

The delivery loop your team already runs, with Kane CLI as the validation engine in the middle.

PlanJira · Confluence · Linear
CodeVS Code · Cursor · Claude Code · Codex
BuildGitHub · GitHub Actions
TestChrome · emulators · simulators
ReleaseTest Manager
Deployyour pipeline
Operateyour infrastructure
MonitorSlack · dashboards
TestMu AI cloudoptional
specs inagents call itPR verdict + checkruns every checkcoverage + evidenceruns remotely
> KANE CLIvalidation engine
connects to Kane CLI optional no line: your stages and tools, unchanged

// Use cases and workflows

What teams can assure with Kane CLI.

Real flows in payments, e-commerce, banking and CRM. Each one is a plain-English sentence and a result Kane CLI has to prove.

Payments

3DS challenge

The 3D Secure step breaks naive automation.

“Pay with a 3DS test card and approve the challenge iframe”

Challenge completes and payment succeeds

Read the use case
E-commerce Found a bug

One-click checkout

Most revenue goes through this path.

“Buy an item with saved address and payment, reach order confirmation”

Order number shown, and the total matches the product page

Read the use case
Banking

Login with 2FA

Every account sits behind this step.

“Log in with the OTP challenge in the test environment”

Dashboard loads with masked accounts

Read the use case
Food delivery

Promo application

A wrong discount costs money on every order.

“Apply a promo and verify the discount math”

Total drops by the promo value

Use case page coming
Travel

Instant book checkout

One failed step loses the booking.

“Complete an instant book with a test payment”

Trip confirmed in the trips tab

Use case page coming
CRM

Record edit through overlays

Loading overlays break automation.

“Edit an account field and save through loading overlays”

Saved value persists after reload

Use case page coming
See the use case library

// From builders

Builders on Kane CLI, in their own words.

𝕏2nd place, Kane CLI Online Hackathon

“My coding agent marked a task done. The chart it built drew the same data twice. So for the @testmuai Kane CLI hackathon I built a gate: Claude Code cannot stop until a real browser agrees. It placed 2nd.”

Abhinav Pangaria

@abhinavpangaria on X · Sep 22, 2026

𝕏Kane CLI Online Hackathon entrant

“for the recorded run, Kane caught a real broken task flow. Elenchos sent that evidence back to Gemini, the agent fixed the implementation, and Kane passed every acceptance criterion on the second attempt”

Zephyr

@Zephyr_cryt on X · Aug 21, 2026

𝕏

“I wasn’t just asking an AI agent to build an e-commerce application and assuming everything worked. I used Kane as part of the process of checking real user flows in the browser.”

Lammy.Sol

@lammy_sol on X · Sep 1, 2026

P

“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.”

Hareesh Vemasani

on Product Hunt · Aug 13, 2026

// Built with Kane CLI

Hackathon projects that run on Kane CLI

Real projects from the Kane CLI Online Hackathon (69 teams, August 2026) and the Kane CLI Hack Day in San Francisco.

demo.studio

$ built with kane-cli

demo.studio

1st place

Give a URL and a goal; get a narrated demo video of the live product. Kane CLI is the hands on the page.

Sam Felix

GuardianKane on GitHub

GuardianKane

2nd place

A verification loop around Claude Code that won't let 'done' mean 'I say it's done.'

Abhinav

KANE vs. ABLE on GitHub

KANE vs. ABLE

3rd place

One AI builds an escape room. Kane must prove it's solvable in a real browser.

Christine Lall

Signed Off on GitHub

Signed Off

Turns a signed freelance scope into a page showing each promise verified in a real browser.

@PhiBao

LENS on GitHub

LENS

AI changes one thing. LENS replays business flows in real Chrome and blocks the agent until they hold.

@Enoch208

mutiny on GitHub

mutiny

A falsification gate: break one promised behavior and flag any Kane test that stays green.

@sneg55

Pactora on GitHub

Pactora

An escrow marketplace. Kane CLI catches the bug the type checker can't; Claude Code patches it.

@mrnetwork0001

Playtest Studio on GitHub

Playtest Studio

Kane CLI Hack Day, SF

Declare a browser game's rules in plain English, and Kane plays it to prove they hold.

@Jiyungi

More Reasons to Love TestMu AI (formerly LambdaTest)

See how TestMu AI improves your testing with seamless integration, quicker results, and unmatched accuracy.

Start free with Google

Users

3M+

Tests

1.5B+

Enterprises

18K+

Countries

132

TestMu AI positioned as a Leader in the 2026 Gartner® Magic Quadrant™

Read Report

TestMu AI recognized in The Forrester Wave™: Autonomous Testing Platforms, Q4 2025

Read Report

Wall of Fame

TestMu AI is the #1 choice for SMBs and enterprises across the globe.

Software review award badges

Enterprise-Grade Security

We safeguard your data and AI systems with global security, privacy, responsible AI, and ESG standards.

Security compliance certification badges

Integrations

Works where you work, 120+ integrations with the tools your team relies on.

Integration partner logos
Capterra
4.6/ 5

540+ reviews

Gartner Peer Insights
4.6/ 5

410+ reviews

Software Advice
4.6/ 5

540+ reviews

As Seen On

Resources and links

Blog

The post that introduced Kane CLI.

Read the blog

Documentation

Install, configure and run Kane CLI in about 2 minutes.

View documentation

GitHub

Browse the source, file issues and follow the roadmap.

Open GitHub

Featured on

Kane CLI, #1 Product of the Day on Product HuntKane CLI listed on Aback LaunchAgentHunterFeatured AI Agent

Frequently asked questions

Kane CLI is an AI CLI testing tool for you and your coding agents: a Web CLI for browser automation and a Mobile CLI for mobile automation. You describe what should happen in plain English, for example "log in as an admin, open the billing page, and verify the plan shows Enterprise", and Kane CLI checks it in a real Chrome browser or your mobile app. You do not write selectors or scripts, and there is no framework to learn.

Every run saves the steps with their screenshots, the console and network logs, and the verdict in a .evidence file. Open it at evidence.lambdatest.com (opening it there uploads nothing), in TestMu AI Test Manager, where runs upload by default, or from the pull request comment the GitHub Action posts.

A pass means every acceptance criterion for that scenario is satisfied. One passing assertion is not enough, and neither is the model's own account of what happened. You write the criteria, and Kane CLI checks whether they hold.

Kane CLI's reasoning decides how to get through the flow. A separate verification step decides the verdict.

Healing has two tiers. Step-heal is the common case: one step's target moved or its markup changed, so Kane CLI re-finds that element by what it is, wherever it now sits. Every other step is untouched. Flow-heal runs only when a step can't be healed at all, and it re-reasons the path, never the acceptance criteria. A wrong total is reported as a bug and is never healed into a pass.

Healing works best when an element's address changes and its meaning stays the same. When a button keeps its label but now does something else, Kane CLI is built to fail the step.

Those are browser automation frameworks you write and maintain by hand, and the maintenance grows with the app. With Kane CLI you describe the flow, it checks it against your live app, keeps passing tests, and repairs them when the UI drifts. When you want code, export any flow to Playwright.

Yes, that's one of the main reasons it exists. Every run supports an --agent flag that switches output to structured NDJSON, so an agent can parse results, read screenshots on failure, and decide what to do next. Point your agent to testmuai.com/kane-cli/agents.md and it learns when and how to call Kane CLI.

Yes. The terminal testing you do locally runs the same way in CI: use --headless and --timeout, and add --agent for structured NDJSON you can parse in a post-step. Kane CLI exits with 0 on pass, 1 on fail, 2 on setup or auth errors, and 3 on timeout, so your pipeline can gate on it.

The CLI is free to install and use. Execution uses your TestMu AI account. CLI testing on your own machine is free, and cloud runs on the TestMu AI grid are billed against your TestMu AI plan.