Kane CLIAGENTS.md

Agent MD file for Kane CLI

This markdown file teaches your coding agent to test in a real browser and on Android and iOS. Save it as AGENTS.md, CLAUDE.md or GEMINI.md, and the agent runs Kane CLI, reads the result and reports pass or fail.

Download a copy for Claude Code or Gemini CLI

Claude Code logo

Claude Code

Rename to CLAUDE.md in your repo

Download
Gemini CLI logo

Gemini CLI

Rename to GEMINI.md in your repo

Download

// Setup

Add the file to your repo, then ask for a test.

Nothing to install besides Kane CLI. The file is plain markdown your agent reads with the rest of the repo.

01

Install Kane CLI

Node 18 or newer. Sign in once with kane-cli login, or with an access key in CI.

02

Add the file

Append it to the instruction file your agent already reads, at the repository root.

03

Ask for a test

Your agent writes the objective, runs Kane CLI, and reports the pass or fail.

you>check that checkout works on staging

// In the file

Each section of the file covers one job your agent does with Kane CLI.

The file is about 200 lines, written for agents to read. Every command and flag in it comes from the Kane CLI skill that TestMu AI maintains.

Test web and mobile apps

Use Kane CLI when a task needs a browser or a mobile app, instead of hand-written Playwright.

Phrase the objective

End every flow in a check, store values by name, and keep each objective under fifteen steps.

Read the result

Parse the NDJSON stream, the final run_end event and the exit code, then report it in plain words.

Save tests that replay

Write _test.md files that record once and replay with no AI cost, then run them as one suite.

Design from requirements

Turn a PRD or a description from chat into designed tests with the assurance commands.

Gate its own pull requests

Run the committed tests before a commit, and stop with the failing step when one of them breaks.

// See it work

Your agent reads the file, then tests the change.

Pick an agent. Each one loads its own file name, runs Kane CLI the way the file says, and reports the pass or fail.

~/shop
reads CLAUDE.md
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
✻ Claude Code~/shop

test the guest checkout

Read CLAUDE.md · Kane CLI rules loaded

Bash Run the guest checkout with Kane CLI

INKANE_CLI_USER_AGENT=claude-code kane-cli run "Go to https://ecommerce-playground.lambdatest.io, check out as a guest, then verify the email and the total." --agent --headless
OUT✓ 1/6 navigate https://ecommerce-playground.lambdatest.io ✓ 2/6 click "Add to cart" ✓ 3/6 fill email = {{email}} ✓ 4/6 click "Continue" ✓ 5/6 assert the email you entered is shown ✓ 6/6 assert the total matches the cart run_end status passed · exit 0 · 41.6s

Guest checkout passed, 6 of 6 steps in 41.6s. The review page shows the email entered, and the total matches the cart. Evidence: evidence.lambdatest.com/r/2f9a1

Ask Claude Code to test a change
passedexit 0

// File or skill

Use the file in any agent, or the skill in Claude Code, Codex CLI and Gemini CLI.

Both teach your agent the same commands, and you can use both in one repo.

AGENTS.md fileany agent

  • Works in any agent that reads an instruction file
  • Lives in the repo, so teammates and CI agents get it too
  • Always in context, about 200 lines
  • You update it by fetching the file again

Kane CLI skillClaude Code · Codex CLI · Gemini CLI

  • Installs for Claude Code, Codex CLI and Gemini CLI in one command
  • Adds a ready check, result cards and a first-run tour
  • Loads detailed references only when a task needs them
  • Saves your preferences between sessions

Frequently asked questions

It is a markdown instruction file for AI coding agents. It tells the agent to use kane-cli for any task that needs a real browser or an Android or iOS app, how to write the objective, how to read the NDJSON result and exit codes, how to save tests that replay, and how to design tests from requirements. Drop it into your repository and the agent follows it on every task.

At the root of your repository. Save it as AGENTS.md for Codex CLI, Cursor, GitHub Copilot, Windsurf, Amp and Jules, as CLAUDE.md for Claude Code, or as GEMINI.md for Gemini CLI. The contents are the same in each, and only the file name changes.

Claude Code reads CLAUDE.md. If your repository already has an AGENTS.md, add the line @AGENTS.md to CLAUDE.md and Claude Code loads it from there, so you keep one copy for every agent.

Not if you use the terminal command on this page. curl -fsSL https://www.testmuai.com/kane-cli-agent/AGENTS.md >> AGENTS.md appends the Kane CLI section to the end of your existing file. If you copy the file by hand, paste it below what you already have.

Use the file when you want one setup that works in any agent and travels with the repository. Use the skill, installed with npx @testmuai/kane-cli-skill, when you work in Claude Code, Codex CLI or Gemini CLI and want the ready check, result cards and references that load on demand. You can use both.

Fetch it again when you update Kane CLI. The file on this page tracks the Kane CLI skill in the LambdaTest/kane-cli repository, and the version it mirrors is written at the top.

Try it on your own app.

Install Kane CLI, add the file, and ask for your first test. Local runs are free.

Start free