Playwright MCP (ExecuteAutomation)
ExecuteAutomation's Playwright MCP server drives a real browser for an agent — navigation, clicking, screenshots, JS execution, and generating Playwright test code.
Add to Claude Desktop
npx -y @executeautomation/playwright-mcp-server
About
This server puts a full Playwright-driven browser behind MCP tool calls: an agent can navigate to a URL, click and fill elements by selector, hover, drag, upload files, read visible text or HTML off the page, and capture screenshots or export a page as PDF. Browser installation is handled automatically, and viewport/device emulation covers over 143 device profiles, so the same tool calls work whether you're testing a desktop layout or an iPhone viewport.
It also has a code-generation mode: start_codegen_session begins recording the actions an agent takes in the browser, and end_codegen_session turns that recording into an actual Playwright test file. That makes it useful for more than one-off scraping or automation — an agent can explore a page interactively and hand you back a runnable test suite instead of just a transcript of what it did.
Key features
- Full browser automation: navigate, click, fill, hover, drag, upload files
- Automatic browser installation, no separate Playwright setup step
- 143+ device emulation profiles for responsive/viewport testing
- Screenshot capture and PDF export of the current page
- JavaScript execution and console log retrieval from the live page
- Codegen sessions that convert recorded actions into runnable Playwright test files
Use cases
- Letting an agent scrape or read data off a page that requires JS rendering
- Automating repetitive browser workflows, like form fills or multi-step checkouts, from natural language
- Generating a Playwright test file by describing the flow to an agent instead of writing selectors by hand
- Cross-device visual QA by driving the same flow across multiple emulated devices
Available tools
Playwright_navigate
Opens a URL with configurable viewport and browser settings.
Playwright_screenshot
Captures the page or a specific element as an image.
Playwright_click
Clicks a page element identified by a CSS selector.
Playwright_fill
Types text into a form field.
Playwright_evaluate
Executes arbitrary JavaScript in the browser and returns the result.
start_codegen_session
Begins recording browser actions so they can be exported as a Playwright test file.
Frequently asked questions
Do I need Playwright already installed to use this?
No — the server handles browser installation automatically, so you don't run a separate playwright install step first.
Can it produce an actual test file, not just perform actions?
Yes — start_codegen_session records the actions taken during a session, and end_codegen_session turns that recording into a runnable Playwright test.