Claude Code Test Runner — Testing skill for Claude Code

Claude Code Test Runner

This project enables full E2E test automation using Claude Code.

Tests are defined using simple natural language steps. Claude Code performs these steps sequentially in a browser through the Playwright MCP, making decisions about element selection, timing, and validation based on the test descriptions.

Why Claude Code as a test runner?

Professional software engineers have been successfully writing automated tests for decades. With the introduction of tools like Claude Code, traditional tests can be written even faster. Why would anyone use Claude Code as a test runner?

First, this test runner was not made to replace traditional unit, integration, or manual testing strategies. It is meant to bolster confidence in the final end-to-end experience of your web application. Ideally, the Claude Code Test Runner sits somewhere between traditional automated E2E tests and manual E2E sanity tests.

Claude Code can execute tests like a real human

Imagine performing manual E2E tests with human-like intuition dozens or hundreds of times each day. That's what Claude Code Test Runner can deliver.

• Natural language test definitions: Tests describe what to test rather than how to test it
• Visual understanding: Can validate UI states based on visual appearance
• Highly scalable: Run as many tests as you are willing to pay for (requires Claude Code subscriptions or incurs API costs).

Claude Code can roll with the punches

E2E tests typically tie together a large number of discrete systems. A lot can happen in the span of a single test, and it is difficult to write traditional E2E tests that account for every edge case.

Claude Code is highly adaptive. It is not tripped up by network blips, minor UX updates, or other innocuous changes and hiccups. It's the perfect test runner for environments where a lot can go wrong.

• Adaptive element selection: Finds elements based on context rather than fixed selectors

...