# playwright

> One API to test and automate Chromium, Firefox and WebKit, for scripts and agents

- Repository: https://github.com/microsoft/playwright
- Category: Browser & computer use
- Stars: 98,322 · Forks: 6,669 · Language: TypeScript · License: Apache-2.0
- Experience: Some experience
- Page: https://ir.adyog.com/repos/microsoft/playwright/

Playwright controls real web browsers from code. You write steps like open a page, click a link, check a heading, and Playwright runs them in Chromium, Firefox or WebKit with the same API. It works on Linux, macOS and Windows, headless or with a visible window.

It comes in several forms. Playwright Test is a test runner with auto-waiting, retrying assertions and a fresh browser context for every test. The library is for plain scripts such as screenshots, PDFs and scraping. A CLI and an MCP server let coding agents and LLM clients drive a browser.

For agents, the README says the CLI is more token-efficient than MCP because it avoids loading large tool schemas and accessibility trees into the model context. The MCP server instead gives agents a structured accessibility snapshot of the page, so no vision model or screenshots are needed. Playwright is also available for Python, .NET and Java.

## What it does

- **Cross-browser, one API**: Runs Chromium, Firefox and WebKit on Linux, macOS and Windows, headless or headed.
- **Auto-wait and web-first assertions**: Playwright waits for elements to be actionable, and assertions retry until they pass, so tests need no artificial timeouts.
- **Isolated tests and Trace Viewer**: Each test gets a fresh browser context, and you can save login state to reuse. Traces capture actions, DOM snapshots, network requests and console messages for debugging.
- **CLI and MCP server for agents**: The CLI targets coding agents like Claude Code and Copilot. The MCP server lets clients such as VS Code, Cursor and Claude Desktop use a browser through accessibility snapshots.
- **VS Code extension**: Run and debug tests from the editor, record tests with CodeGen, and pick locators by hovering over elements.

## What you could build

- Write end-to-end tests for a web app and run them in parallel across three browsers.
- Capture screenshots or generate PDFs of web pages with a short script.
- Emulate a mobile device such as an iPhone 15 to check how a page looks.
- Let a coding agent test a web flow and take screenshots using the CLI.
- Connect an MCP client so an AI assistant can fill forms and navigate sites.

## Who it's for

Suits web developers, testers and people building browser-using AI agents; you should know basic JavaScript or TypeScript and have Node.js installed.

## Get started

For testing, run the init command to set up a project with example tests, then run them. For AI agents, add the MCP server to your client's config or install the CLI as the README describes.

```
npm init playwright@latest
```
```
npx playwright test
```
```
npx @playwright/mcp@latest
```

Tags: playwright, browser-automation, e2e-testing, mcp, web-scraping