Chrome MCP Server: the Chrome extension that lets AI control the browser for automation

Chrome MCP Server is a Chrome-based extension. It allows AI assistants (e.g., Claude) to directly control the user's Chrome browser through the Model Context Protocol (MCP) protocol, enabling complex automated operations, content analysis, and semantic search. Compared to traditional browser automation tools (e.g., Playwright), it does not need to start a separate browser process and directly utilizes the user's open Chrome browser, preserving the user's login status, settings and habits. It runs completely natively, protects user privacy, supports streaming HTTP connections and cross-tab contexts, and offers more than 20 utilities. Users need to install the Chrome extension andmcp-chrome-bridgeSoftware package. It is developed by hangwin, follows the MIT license, and is suitable for users who wish to improve the efficiency of their browsers with AI.

Function List

• Browser automation: supports AI assistants to control the browser and perform actions such as clicking, typing, and navigating.
• Semantic Search: Built-in vector database to intelligently search browser tab content.
• Content analysis: AI-driven text extraction and similarity matching for fast parsing of web information.
• Screenshot tool: support webpage screenshot, optimized base64 image compression to save resources.
• Network Monitoring: Monitor network requests in real time and analyze web page loading performance.
• Interactive Element Detection: Enhanced tool to recognize interactive elements, including div elements.
• Console output capture: viachrome_consoletool to get browser console logs.
• Bookmarks and history management: Support AI to manipulate user bookmarks and browsing history.
• Cross Tab Context: Enables data sharing and manipulation across multiple tabs.
• Local operation: All operations are done locally to ensure data privacy.
• Streaming HTTP connection: provides efficient communication with fast response time.
• SIMD Acceleration: WebAssembly optimizes vector arithmetic for a 4-8x speedup.
• Script injection: supports dynamic injection of JavaScript scripts to extend functionality.

Using Help

Installation process

To use Chrome MCP Server, users need to complete the following installation steps. The whole process is simple, but you need to make sure that your environment is properly configured.

1. Inspection of environmental requirements
   • Make sure Node.js is installed (version 18.19.0 or above).
   • Install pnpm or npm as a package manager.
   • Use Chrome or Chromium browser (latest version recommended).
2. Install Chrome Extension
   • Visit the GitHub releases page at https://github.com/hangwin/mcp-chrome/releases.
   • Download the latest version of the Chrome extension (.crx(Documentation).
   • Open Chrome and go tochrome://extensions/The
   • Enable "Developer Mode" (switch in the upper right corner).
   • drag-and-drop.crxGo to the Extensions page and click "Add Extension" to complete the installation.
3. mountingmcp-chrome-bridge
   • Open a terminal and run the following command to install globallymcp-chrome-bridge::

     npm install -g mcp-chrome-bridge
     

     Or use pnpm:

     pnpm install -g mcp-chrome-bridge
     

   • If using pnpm v7+, the postinstall script is disabled by default, you need to enable the script:

     pnpm config set enable-pre-post-scripts true
     

   • If automatic registration fails, register manually:

     mcp-chrome-bridge register
     

4. Configuring the MCP Client
   • Using CherryStudio as an example, modify the MCP client configuration file to add the following:

     {
     "mcpServers": {
     "chrome-mcp-server": {
     "type": "streamableHttp",
     "url": "http://127.0.0.1:12306/mcp"
     }
     }
     }
     

   • If the client only supports stdio connections, you need to check that themcp-chrome-bridgeInstallation path and configure the stdio connection method (refer to the official documentation)CONTRIBUTING.md).
5. Starting services
   • Once the installation is complete, the Chrome extension is automatically activated.
   • (of a computer) runmcp-chrome-bridgeto ensure that the service listens on thehttp://127.0.0.1:12306/mcpThe
   • An AI assistant (e.g. Claude) can then connect to the browser via the MCP protocol.

Main Functions

Here are a few features of Chrome MCP Server and the detailed procedure.

Browser Automation

• Functional Description: The AI assistant can control the browser via the MCP protocol to perform actions such as clicking, typing, and navigating. For example, auto-filling forms or batch opening web pages.
• procedure::
  1. Ensure that the MCP service is started and the AI assistant is connected.
  2. Enter commands in the AI Assistant interface, such as "Open Google and search for the keyword 'AI'".
  3. Chrome performs the action automatically, and AI parses the content of the page and returns the result.
  4. Complex tasks, such as "search for products on an e-commerce site and add them to the shopping cart," can be specified through commands.

semantic search

• Functional Description: Using the built-in vector database, AI can intelligently search the contents of browser tabs to quickly locate relevant information.
• procedure::
  1. Open multiple tabs containing the content to be searched.
  2. Enter a command in the AI Assistant, such as "Find all tabs that mention 'AI technology'".
  3. The AI analyzes the tab content and returns matching results, including titles and summaries.
  4. The AI can be further instructed to extract specific passages or jump to relevant tabs.

content analysis

• Functional Description: AI can extract web page text for similarity matching or summarization. For example, analyzing key points of a news article.
• procedure::
  1. Open the target web page, such as a news article.
  2. Enter commands into the AI assistant, such as "Summarize the main content of the current web page".
  3. The AI will extract the text and generate a concise summary or list of keywords.
  4. The AI may be asked to compare the content of multiple web pages to identify commonalities or differences.

Screenshot Tools

• Functional Description: Support webpage screenshot, optimized base64 image compression to reduce resource usage.
• procedure::
  1. Open the web page you need to take a screenshot of.
  2. Input commands in AI assistant, such as "capture full screen image of current webpage".
  3. The AI generates screenshots and returns base64 encoded images.
  4. The image can be saved or the AI can be asked to analyze the content of the image (e.g., extract text).

network monitoring

• Functional Description: Monitor web requests for pages in real time to analyze loading performance or troubleshoot problems.
• procedure::
  1. Open the target web page.
  2. Enter a command into the AI assistant, such as "Monitor web requests for the current web page".
  3. The AI returns a list of requests, including URLs, status codes, and response times.
  4. Slow loading resources or unusual requests can be further analyzed.

caveat

• Ensure that Chrome extensions andmcp-chrome-bridgeVersion consistency to avoid compatibility issues.
• Running locally may take up some memory, it is recommended to close unnecessary tabs.
• Check out GitHub'sCONTRIBUTING.mdGet more configuration details or contribute code.

application scenario

1. automated office work
   Employees can use AI assistants to batch process web tasks such as auto-filling reports, extracting data, or monitoring website updates.Chrome MCP Server utilizes the existing browser environment and saves time by eliminating the need to re-login.
2. Content research
   Researchers can quickly organize information from multiple web pages with semantic search and content analysis features. For example, analyze industry reports or academic articles to extract key points and generate summaries.
3. Web development and testing
   Developers can utilize network monitoring and console output capture to debug web page performance or automate test interaction flow. Script injection tools support dynamic modification of web pages for easy prototyping.
4. Personal Efficiency Improvement
   Users can let AI manage bookmarks, browsing history or perform repetitive operations such as checking mailboxes regularly or downloading files in batches to enhance daily efficiency.

QA

1. Does Chrome MCP Server require an internet connection?
   It runs completely locally, and the communication between the AI assistant and the browser is done locally, without the need for an internet connection. However, the AI assistant itself may require networking, depending on the AI model used.
2. What AI assistants are supported?
   Currently supports AI assistants compatible with the MCP protocol such as Claude. More models may be extended in the future, check GitHub for updates.
3. How do you ensure privacy and security?
   All operations are performed locally and data is not uploaded to the cloud. It is recommended to download extensions from the official GitHub to avoid third-party sources.
4. What if the installation fails?
   Check that the Node.js and pnpm versions meet the requirements, and make sure that Chrome developer mode is enabled. Refer toCONTRIBUTING.mdOr submit an Issue for help.