chrome-bridge-mcp
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| CHROME_BRIDGE_PORT | No | Port for Chrome DevTools Protocol connection | 9222 |
| CHROME_BRIDGE_RUNTIME | No | Runtime directory for Chrome profile and logs | $HOME/runtime/.chrome-bridge-mcp |
| CHROME_BRIDGE_AUTO_START | No | Auto-start Chrome if not running | 1 |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| chrome_statusB | Check whether a Chrome DevTools Protocol endpoint is reachable. |
| list_tabsB | List open Chrome tabs exposed by the DevTools endpoint. |
| open_urlB | Open a URL in a new Chrome tab. |
| google_searchC | Open a Google search query in Chrome. |
| search_google_and_extractA | Search Google in Chrome and return structured results. If login/CAPTCHA is detected, returns needsUser=true with tabId for manual continuation. |
| open_google_aiC | Open Google AI Mode with an initial prompt and read the visible AI conversation state. |
| google_ai_askB | Send a follow-up prompt to an existing Google AI Mode tab and read the updated conversation. |
| google_ai_readA | Read visible text and detected state from a Google AI Mode tab. |
| start_google_ai_sessionB | Start a persisted Google AI Mode conversation session for chat-through-Codex/Claude workflows. |
| continue_google_ai_sessionA | Send a follow-up message to a persisted Google AI conversation session and return the updated visible conversation. Defaults to the active session. |
| read_google_ai_sessionA | Read persisted metadata and current visible text for a Google AI conversation session. Defaults to the active session. |
| end_google_ai_sessionB | End a Google AI conversation session: export it, close the Chrome tab, and clear the active session if it matches. |
| export_google_ai_sessionA | Export a persisted Google AI conversation session to Markdown in the project reports directory. Defaults to the active session. |
| get_current_pageB | Extract title, URL, selected text, and visible page text from a Chrome tab. |
| get_selectionC | Return the selected text from a Chrome tab. |
| extract_google_resultsA | Extract structured organic search results from a Google search results page. |
| detect_human_interventionB | Detect whether the current page appears to require login, CAPTCHA, consent, or manual verification. |
| fill_textB | Fill an input, textarea, or contenteditable element. |
| click_selectorB | Click an element by CSS selector. |
| ask_ai_pageA | Submit a prompt to a generic browser AI page and read the response. Login/CAPTCHA pages return needsUser=true. |
| run_jsB | Run JavaScript in a Chrome tab and return the JSON-serializable result. |
| wait_for_userB | Return instructions for the human to complete login, CAPTCHA, consent, or other browser-only actions. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 22 tools
There are multiple tools for interacting with Google AI (e.g., ask_ai_page, google_ai_ask, open_google_ai) that could confuse an agent about which to use. Also, detect_human_intervention and wait_for_user serve similar purposes. However, most tools for general browser control are distinct.
Most tools follow a verb_noun pattern (e.g., click_selector, open_url), but several start with the noun (e.g., google_ai_ask, google_search) breaking consistency. Additionally, names like search_google_and_extract are overly verbose.
With 22 tools, the server covers both general browser automation and Google AI interactions. This is slightly high but still reasonable; each tool has a clear purpose, though some duplication could be consolidated.
The general browser tools cover core actions like open, click, fill, and get page info. Google AI session management is thorough. However, missing tools for closing tabs, keyboard input, or screenshots create minor gaps.