MCP server for controlling Chrome browser via native messaging or WebSocket, enabling automated browser interactions including navigation, clicking, typing, screenshots, and content extraction
Chrome Pilot has 21 tools with complete input schemas and descriptions, but quality is inconsistent. Naming follows verb_noun convention well (navigate, click, screenshot). However, descriptions are often generic (e.g., 'Click on an element in the page', 10 words, lacks context on when to use vs alternatives). Parameter descriptions exist but lack constraints: 'selector' appears 12 times with identical minimal descriptions, no guidance on CSS vs XPath trade-offs. Output schemas are not documented, responses are inferred from code (success flag, message field). Error handling is present but generic ('Click failed: ...'). No tool annotations (readOnlyHint, destructiveHint). Critical gap: 'click_advanced' has no required parameters, inviting misuse. 'search_elements' conflates textQuery and selector without documenting mutual exclusivity.
Clear the value of an input element
Click on an element in the page
Advanced click with options for selector, coordinates, navigation waiting, and timeout
Click at specific coordinates on the page
Close a browser tab
Create a new browser tab with a specified URL
Fill an input element with a value
Get the content of the page or a specific element
Output schemas not documented. Responses inferred from code (success, message, data fields). LLMs cannot plan downstream calls without knowing return structure.
Parameter descriptions lack constraints and context. 'selector' appears 12 times with identical minimal text. No guidance on CSS vs XPath, required format, or when to use each.
click_advanced has no required parameters (empty required array). Invites misuse, LLM may call with no selector or coordinates, causing silent failures.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot distinguish safe reads from destructive writes without explicit hints.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
Get information about the currently active tab
Get HTML content of the page or a specific element
Get all interactive elements on the page (buttons, inputs, links, etc.)
Get all open browser tabs
Get text content of the page or a specific element
Navigate to a URL in the browser
Take a screenshot of the current page
Take a screenshot of a specific element
Take a full page screenshot
Scroll the page in a specified direction
Search for elements by text or selector with optional coordinate inclusion
Type text into an input field
Wait for an element to appear on the page
search_elements conflates textQuery and selector without documenting mutual exclusivity. LLM may pass both or neither, causing ambiguous behavior.