MCP SERVERGives your AI assistant live page context and browser control for writing and debugging Playwright tests.
playwright-mcp is a local MCP server that lets your AI assistant inspect and control a Playwright browser. It supplies page structure, screenshots and interactive-element context to help write and debug tests against the actual interface. You can also use it to extract page text or automate browser interactions.
The server can run custom Playwright JavaScript against the current page, so it can perform actions that change website data.
When you need Playwright tests for a page your assistant has not seen, give it the page’s structure and interactive elements. It can use that context to write selectors and test code based on the actual interface.
Toolsinit-browserget-full-domget-interactive-snapshot
When a test behaves differently from what you expect, the assistant can inspect screenshots and page content, then run targeted Playwright code. Returned values, console logs and errors give it evidence to help diagnose the failure.
Toolsinit-browserget-screenshotget-full-snapshotexecute-code
For a browser task that involves several clicks and text entries, the assistant can identify buttons, links and inputs, then write and run Playwright code to carry out the sequence.
Toolsinit-browserget-interactive-snapshotexecute-code
When you want information from a web page, the assistant can retrieve its headings, paragraphs and lists. That gives it the text to summarize or analyze without requiring you to copy the content into the conversation.
Toolsinit-browserget-text-snapshot
Showing all 7
execute-codeRuns custom Playwright JavaScript against the current page for automation or debugging. It can interact with the page, so its effects depend on the code executed.
init-browserInitializes a browser and navigates to a supplied URL.
get-full-domRetrieves the current page’s full DOM to inspect its HTML structure. Deprecated in favor of get-context, which is absent from the collected tool list.
get-full-snapshotCaptures the page’s visible content, including text, images and forms, to provide context for understanding the page as a whole.
get-interactive-snapshotShows buttons, links and inputs with an annotated screenshot to help identify targets for browser automation.
get-screenshotCaptures a screenshot of the current page to inspect its visual appearance.
get-text-snapshotExtracts the page’s text, including headings, paragraphs and lists, for reading or content extraction.
No matching results.
The server runs locally and needs Node.js, Playwright and its Chromium browser; connect it through an AI tool that supports MCP.
Have Node.js, Playwright and its Chromium browser installed, with npx available on PATH.
Save this configuration in your project's .cursor/mcp.json. If the file already contains other servers, add the playwright-mcp entry under its existing mcpServers object.
.cursor/mcp.jsonin your projectJSON{
"mcpServers": {
"playwright-mcp": {
"command": "npx",
"args": [
"-y",
"playwright-mcp"
]
}
}
}Check the result
Check that Cursor sees the tools
Look for the server's tools under Available Tools. If the server is disabled, enable it from Customize in the sidebar.