Autospec is an open-source AI agent that takes a web app URL and autonomously QAs it, and saves its passing specs as E2E test code

autospec

Open source end-to-end (e2e) test generation for web apps

Autospec is an AI agent that autonomously explores your web application, generates commonsense test specifications, and executes them — producing reusable Playwright test files. It uses vision and language models to mimic real user judgement on the entire UI after each interaction, deciding whether behavior is correct rather than checking for regressions against rigidly defined previous behavior.

• Tests new features immediately after implementation, not just regressions.
• Requires no configuration — point it at a URL and go.
• Generates standard Playwright .spec.js files you can re-run anytime.

Quick start

Generate and run 10 specs on TodoMVC:

npx autospecai --url https://todomvc.com/examples/react/dist/ --apikey YOUR_ANTHROPIC_API_KEY

You'll need to say "yes" to install the autospecai package, and the first run may take a few minutes to download dependencies like browser binaries.

When the run completes, you'll see a summary of passed and failed tests. Passing specs are saved as Playwright test files in the trajectories/ folder alongside video recordings and screenshots. Re-run them anytime:

npx playwright test

Depending on your existing Playwright setup, you may need to add "trajectories" to the testDir in your playwright.config.js file.

Using environment variables instead of passing keys as a flag

Copy the sample .env file and fill in the API key for your chosen model:

cp .env.example .env
nano .env

Configuration

> npx autospecai --help
    Usage: npx autospecai --url <url> [--model <model>] [--spec_limit <limit>] [--help | -h]

    Required:
    --url <url>          The target URL to run the autospec tests against.

    Optional:
    --help, -h           Show this help message.
    --version, -v        Show ver