Set Up Playwright
Give your agent full browser automation — headless Chromium, page navigation, form filling via Playwright.
Give your agent full browser automation with Playwright. Unlike agent-browser (which uses accessibility tree snapshots), Playwright gives your agent direct control over a headless Chromium instance — click elements, fill forms, take screenshots, and run end-to-end workflows.
1. Install
crabtalk hub install microsoft/playwrightThis installs the Playwright MCP server and downloads Chromium. The setup runs npx playwright install chromium automatically.
2. Verify
crabtalk attach> Go to crabtalk.ai and take a screenshotThe agent launches a headless browser, navigates to the page, and returns a screenshot.
When to use Playwright vs. agent-browser
| Use case | Tool |
|---|---|
| Quick page reading, content extraction | agent-browser |
| Complex multi-step browser workflows | Playwright |
| Form filling and clicking specific elements | Both work |
| Screenshots and visual verification | Playwright |
| Token-efficient page summarization | agent-browser |
agent-browser is lighter and more token-efficient for simple reads. Playwright is more powerful for complex automation and testing workflows.
What's next
- Use Browser Tools — the lighter alternative with agent-browser
- MCP Servers — how tool servers connect to the daemon
- Hub — discover and install more packages