Home / MCP / BrowserTools MCP Server
Provides browser monitoring and auditing tools for MCP clients via a local Chrome extension and Node server.
Configuration
View docs{
"mcpServers": {
"browser_tools_mcp": {
"command": "npx",
"args": [
"@agentdeskai/browser-tools-mcp@latest"
]
}
}
}BrowserTools MCP lets you run a local MCP server that exposes browser monitoring and interaction capabilities to AI clients inside your IDE. It captures logs, screenshots, network data, and DOM information through a Chrome extension, all kept on your machine for privacy and control, enabling powerful browser-aware automation and analysis from MCP-enabled agents.
You connect an MCP client (such as Cursor or another compatible editor) to the BrowserTools MCP server to collect browser data and run audits. Start the two required servers locally, then enable the Chrome extension and open the BrowserTools MCP panel in your browser tools. Use the MCP client to issue audit and debugger commands that run against the current page and return structured results you can act on from your AI workflow.
Audits cover Accessibility, Performance, SEO, and Best Practices, with specialized modes like Audit Mode to run all audits in sequence and Debugger Mode to execute a predefined debugging workflow. All logs stay on your machine and are not sent to external services by default.
# Install the MCP client integration into your IDE
npx @agentdeskai/browser-tools-mcp@latest
# Install the local BrowserTools server
npx @agentdeskai/browser-tools-server@latestTwo components are required for a complete setup and they work together to enable MCP access to browser data. The MCP client communicates with the local Node server, which in turn talks to the Chrome extension to collect logs and capture screenshots. Your setup stores logs locally on your machine, and you can wipe logs as needed from the MCP server.
Before using audits, ensure you have an active browser tab and the BrowserTools extension enabled. You can trigger audits from your MCP client by asking for accessibility, performance, SEO, or best-practices checks, or by entering the dedicated modes that run multiple tools in sequence.
If things arenβt working, quit all Chrome windows, restart the local browser-tools-server, and ensure only one instance of Chrome DevTools is open. Reopen the BrowserTools MCP panel in the Chrome DevTools, and retry the client connection.
If issues persist, restart the MCP client, check that the MCP server processes are running, and verify there are no conflicting network ports between the server and extension.
WCAG-focused checks for color contrast, missing alt text, keyboard navigation traps, ARIA attributes, and other accessibility concerns.
Lighthouse-driven analysis of render-blocking resources, DOM size, image optimization, and other factors impacting load times.
Evaluates on-page SEO elements such as metadata, headings, links, and structure to suggest improvements for search visibility.
General web development best practices checks to improve reliability and security.
NextJS-specific prompts to assess SEO and best-practices within NextJS applications.
Runs all auditing tools in a predefined sequence, optionally including a NextJS audit if applicable.
Executes a sequence of debugging tools to help diagnose issues in the current page.