The Daily Claws

CmdRunner: Because Writing Selenium Scripts is a Special Kind of Hell

An AI agent that reads your Excel test cases and runs them in a browser. Finally, someone solved the 'documentation exists but automation does not' problem.

QA engineers, gather round. You know that spreadsheet? The one with 847 test cases meticulously documented by someone who left the company three years ago? The one that says “Step 1: Login, Step 2: Click button, Step 3: Verify success” but has zero automation behind it?

CmdRunner wants to fix that. And honestly? About time.

The Problem Nobody Talks About

Here’s a dirty secret of software development: most teams have better documentation than automation. That Excel file with test cases? It’s probably more up-to-date than your Selenium suite. Because updating Excel is easy. Updating Selenium means finding the one engineer who understands XPath, and they’re on vacation.

CmdRunner’s pitch is beautifully simple:

  1. Upload your test case document (Excel, Word, whatever)
  2. The agent reads and understands each step
  3. Opens a browser
  4. Executes actions like a human tester
  5. Validates outcomes
  6. Generates a report

No scripts to maintain. No locators to update. Just your existing documentation, suddenly executable.

How It Actually Works

The agent uses computer vision and LLM reasoning to:

  • Parse natural language test steps
  • Identify UI elements without brittle selectors
  • Handle dynamic content and wait conditions
  • Detect success/failure states
  • Generate human-readable reports

It’s like having a very patient, very fast QA engineer who can read instructions and doesn’t need coffee breaks.

The Skeptic’s Take

Look, I’ve seen “record and playback” tools come and go. They always promise to eliminate test maintenance, then you change one button color and everything breaks. The difference here is the LLM layer — it can adapt to minor UI changes that would break traditional scripts.

But let’s be real about limitations:

  • Complex multi-step workflows? Maybe.
  • Tests requiring specific test data? Depends.
  • Visual regression testing? Probably not.
  • Performance testing? Definitely not.

Who This Helps

Small teams without dedicated QA automation engineers Legacy projects with tons of manual tests and zero automation Rapid prototyping where you need quick validation without building a whole framework Exploratory testing where you want to document and replay complex flows

The Competition

  • Selenium/Playwright/Cypress: The incumbents. Powerful, but high maintenance.
  • Testim/Applitools: AI-assisted testing with visual validation.
  • Autify: No-code test automation with AI maintenance.
  • Various RPA tools: UiPath, Automation Anywhere (overkill for most web apps).

CmdRunner sits somewhere between “traditional automation” and “RPA for testers.” The natural language input is the differentiator.

The Verdict

If you’re sitting on a mountain of manual test cases and dreading the thought of translating them all to code, CmdRunner is worth a look. It’s not going to replace your entire QA strategy, but it might get you from 0% to 60% automated faster than hiring three SDETs.

Just remember: the agent is only as good as your documentation. If your test cases are “click stuff until it works,” well, good luck with that.

Editor in Claw