Selenium

You run an MCP server that orchestrates browser automation through Selenium, with stealth features, multi‑browser support, and robust session management to automate web tasks reliably across environments.

How to use

You start by launching the server in development mode or via the MCP CLI, then you create a browser session, navigate to pages, and interact with elements just like you would with Selenium. Use stealth options when you need to evade basic bot detection, and enable human‑like delays and scrolling for more realistic automation. You can manage multiple sessions at once, store cookies, and take screenshots for debugging.

How to install

Prerequisites you need before installing are Python 3.8 or higher and a browser installed (Chrome and/or Firefox). You also need basic system dependencies for Selenium.

# Clone or download the server package
# Example folder name shown here; use your actual path
cd mcp_selenium_server

# Install Python dependencies
pip install -r requirements.txt

Configuration and usage notes

The server supports predefined configurations for different scenarios: a default setup, a stealth mode tuned for evasion, and a performance‑oriented profile. You can also configure proxy settings to route traffic through HTTP, HTTPS, or SOCKS proxies. When using stealth, enable undetected chrome and randomized user agents and viewports to reduce detection risk. Use human‑like delays and scrolling to mimic real user interactions.

# Default configuration snippet
DEFAULT_CONFIG = ServerConfig(
    max_sessions=10,
    session_timeout=3600,
    default_browser="chrome",
    browser_options=BrowserOptions(
        headless=False,
        window_width=1920,
        window_height=1080,
        incognito=True
    )
)

# Stealth configuration snippet
STEALTH_CONFIG = ServerConfig(
    browser_options=BrowserOptions(
        headless=True,
        user_agent="Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36...",
        disable_images=True,
        custom_args=[
            "--disable-blink-features=AutomationControlled",
            "--disable-dev-shm-usage",
            "--no-first-run"
        ]
    ),
    detection_evasion=DetectionEvasionConfig(
        use_undetected_chrome=True,
        stealth_mode=True,
        randomize_user_agent=True,
        randomize_viewport=True
    )
)

Usage patterns

• Start a browser session and choose the desired options. The final start command can be a Python invocation such as python server.py in development mode. If you are using the MCP CLI, you can initiate your server with mcp dev server.py or install with mcp install server.py to prepare the environment for usage.

• From a client, create a session, then perform navigation, element interaction, keyboard input, and file uploads. Capture screenshots for verification, and close the browser when your workflow completes.

Examples of common workflows

# Start a browser session
result = start_browser(
    browser_type="chrome",
    options={"headless": False, "window_width": 1920, "window_height": 1080},
    detection_evasion={"use_undetected_chrome": True, "stealth_mode": True}
)
session_id = result["session_id"]

# Navigate to a URL
navigate_to_url(session_id, "https://example.com")

# Find and interact with elements
element = find_element(session_id=session_id, strategy="css_selector", value="input[name='search']")
type_text(session_id=session_id, strategy="css_selector", value="input[name='search']", text="selenium automation")
click_element(session_id=session_id, strategy="css_selector", value="button[type='submit']")

# Take a screenshot
take_screenshot(session_id=session_id, file_path="/path/to/screenshot.png")

# Close the session
close_browser(session_id)

Available tools

start_browser

Starts a new browser session with the specified browser type and options.

close_browser

Closes a specific browser session by its session_id.

detect_available_browsers

Detects browsers installed on the host and returns a list of available options.

get_recommended_browser

Returns the browser recommended for the current environment.

check_browser_support

Verifies if the requested WebDriver is supported by the server.

navigate_to_url

Navigates the active browser session to a specified URL.

get_page_info

Retrieves information about the current page, such as title and URL.

execute_script

Executes a JavaScript snippet in the context of the active page.

find_element

Finds elements using various strategies (id, css_selector, xpath, etc.). Supports multiple elements.

click_element

Clicks on a located element.

type_text

Types text into an input or editable element.

upload_file

Uploads a file to an input of type file.

perform_mouse_action

Performs advanced mouse actions like hover, drag and drop, right-click, and double-click.

send_keys

Sends special keys (Enter, Tab, arrows, etc.) to the focused element.

take_screenshot

Takes a full page or element screenshot and saves it to a file.

apply_stealth_mode

Applies stealth techniques to reduce automation detection.

randomize_user_agent

Switches to a random user agent for the browser session.

randomize_viewport

Changes the browser viewport size to emulate different devices.

simulate_human_typing

Types with realistic typing delays to mimic human input.

scroll_like_human

Scrolls gradually to imitate natural user scrolling.

add_random_delay

Introduces random delays between actions to simulate human pacing.

list_active_sessions

Returns a list of currently active browser sessions.

get_session_info

Provides detailed information about a specific session.

close_all_sessions

Closes all active sessions.

cleanup_expired_sessions

Removes sessions that have exceeded their timeout.

manage_cookies

Manages cookies for the current session, including adding and removing cookies.

get_server_status

Returns the current status of the MCP server.

update_server_config

Updates server configuration using presets or custom values.