Selenium MCP Server (Console-based)

A console-based Model Context Protocol (MCP) server for Selenium WebDriver, compatible with IDEs and desktop AI agents. Communicates via stdin/stdout (no HTTP server required).

Features

• Multi-session browser management
• All major browser actions (navigation, click, type, clear, screenshot, etc.)
• Element queries (find, get text/attribute/property/tag, isDisplayed, isEnabled)
• Frame and window switching
• Script execution (sync/async)
• File upload
• Browser navigation (back, forward, refresh)
• Element hover and right-click
• Page info (source, title, URL)
• Cookie management (get, set, delete, getAll)
• Alert/dialog handling (accept, dismiss, getText, sendKeys)
• Window resizing and maximizing
• Element screenshot
• Wait for conditions (title, url, element state)
• Custom wait (sleep)
• Logging (for debugging)

Usage

1. Install dependencies

npm install

2. Start the server

npm start

3. Communicate via stdin/stdout

Send JSON commands (one per line) to stdin. Read JSON responses from stdout.

Example (PowerShell):

node src/server.js

Example command (start browser):

{"action": "start_browser", "browser": "chrome"}

Example command (navigate):

{"action": "navigate", "url": "https://example.com"}

Example command (find element):

{"action": "find_element", "by": "css", "value": "h1"}

Example command (click):

{"action": "click", "by": "css", "value": "#submit"}

Example command (get cookies):

{"action": "get_cookies"}

Example command (accept alert):

{"action": "accept_alert"}

Example command (execute script):

{"action": "execute_script", "script": "return document.title;"}

4. Supported Actions

Send { "action": "help" } to get a list of all supported actions.

5. Sessions

• Each browser session gets a unique sessionId.
• Pass sessionId in commands to target a specific session.
• Use switch_session to change the default session.

6. Error Handling

• All responses have a status field: ok or error.
• On error, a message field is included.

Notes

• Requires ChromeDriver/GeckoDriver/etc. in PATH for browser automation.
• For advanced scripting, see Selenium WebDriver JS API.

MIT License