Browser automation plugins for the Cursor IDE Marketplace. Control a real Chrome browser through natural language -- navigate, click, type, extract data, and take screenshots.
| Plugin | Description |
|---|---|
| browse | Automate browser interactions via MCP tools. Navigate pages, fill forms, extract data, take screenshots. No API key needed for local mode. |
| functions | Deploy serverless browser automation to Browserbase cloud using the bb CLI. |
cd plugins/browse
npm install # Installs dependencies and auto-builds TypeScriptThe MCP server starts automatically when the plugin is active in Cursor. Just ask:
- "Go to Hacker News and get the top 5 stories"
- "Fill out the signup form on example.com"
- "Take a screenshot of localhost:3000"
For stealth browsing, proxies, and CAPTCHA solving:
export BROWSERBASE_API_KEY="your-api-key"
export BROWSERBASE_PROJECT_ID="your-project-id"Get credentials at browserbase.com/settings.
The browse plugin runs an MCP server over stdio that wraps Playwright:
Cursor Model ──MCP tool call──▶ MCP Server ──Playwright──▶ Chrome
▲ │
│ │
└──── screenshot file path ──────┘
└──── interactive elements ──────┘
- No API key needed for local Chrome automation
- Headed browser -- you can watch the automation happen
- 10 MCP tools: navigate, click, type, snapshot, screenshot, scroll, evaluate, select, wait, close
- Persistent sessions -- cookies and login state carry over via Chrome profile
node scripts/validate-template.mjs- macOS/Windows: Install from google.com/chrome
- Linux:
sudo apt install google-chrome-stable
rm -rf plugins/browse/.chrome-profile