MCP Server for browser automation via Selenium WebDriver + geckodriver (Firefox)
Server has 45 tools with consistent naming (browser_* prefix) and descriptions present for all. However, descriptions are often terse (10-30 chars), parameter schemas lack depth (missing min/max bounds, enums where appropriate), and output schemas are not documented. Error handling is minimal, no recovery guidance or actionable error messages. Tool composition is reasonable (single responsibility per tool), but parameter descriptions are sparse. No evidence of input validation or constraint documentation in descriptions.
Accept a browser dialog (alert, confirm, prompt)
Navigate back in browser history
Clear browser storage (localStorage/sessionStorage)
Click an element on the page by CSS selector
Close the current browser session
Delete a cookie from the browser
Dismiss a browser dialog (alert, confirm, prompt)
Double-click an element
Parameter descriptions are sparse and lack constraint documentation. E.g., browser_press_key accepts 'key' but does not enumerate valid values (enter, tab, escape, etc.) in the description; browser_get_network_log accepts 'slow_ms' with no min/max bounds documented.
Output schemas are not documented. Tools like browser_get_network_log, browser_get_console_logs, browser_get_accessibility_tree return complex structures but no schema is visible in the source code. LLMs cannot plan downstream operations without knowing response structure.
Error handling is absent. No evidence of try-catch blocks, validation, or recovery guidance in tool implementations. Tools like browser_click, browser_fill, browser_navigate will fail silently or throw raw exceptions without actionable error messages.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | <=2025-11-25 | v2 |
Drag an element and drop it on another element
Execute arbitrary JavaScript in the page context
Fill a form field with text
Find all elements matching a CSS selector
Navigate forward in browser history
Get accessibility snapshot of the page or an element
Get computed CSS style of an element
Get console output (log, warn, error, info, debug messages)
Get all cookies from the browser
Get an attribute value from an element
Get the HTML of an element
Get the location (x, y coordinates) of an element
Get the size of an element
Get the text content of an element
Get JavaScript errors that occurred on the page
Get network requests and responses
Get the full HTML source of the current page
Get values from browser storage (localStorage/sessionStorage)
Get the title of the current page
Get the current URL of the browser
Hover over an element
Check if an element is enabled
Check if an element is visible on the page
Navigate to a URL in the current browser session
Open a new browser session with Firefox via geckodriver
Press a keyboard key
Refresh the current page
Right-click an element
Take a full-page screenshot of the current page
Scroll the page or an element
Select an option from a dropdown/select element
Set a cookie in the browser
Set a value in browser storage (localStorage/sessionStorage)
Get the current status of the browser session
Upload a file to a file input element
Wait for an element to appear on the page
Wait for a page navigation to complete
Tool descriptions are terse (10-50 chars) and lack context for LLM selection. E.g., 'Close the current browser session' does not explain when to call it vs browser_navigate, or what happens to open tabs. Descriptions should be 50-200 chars with WHEN/WHY guidance.
No input validation or constraint enforcement visible. Parameters like 'selector' (CSS selector) and 'script' (JavaScript) accept arbitrary strings with no sanitization against injection attacks. Untrusted agent input should be validated before execution.