mcp -rag-web-browser

This MCP server exposes the RAG Web Browser Actor via the Model Context Protocol (MCP). It enables AI agents and language models to perform web interactions through a lightweight local server: conducting web searches, scraping and extracting content from top results, and fetching specific URLs. The server communicates over standard input/output (stdio), making it compatible with common MCP clients and enabling streaming content in real time when used with compatible tooling. The included tool is named search and supports both Google-style queries and direct URL fetches, returning cleaned content in Markdown (with options to customize output format). You can run this locally to prototype browsing-enabled workflows or to integrate web data into your agent pipelines during Standby mode in MCP workflows.

Prerequisites:

• Node.js v18 or higher
• npm (comes with Node.js)
• An Apify API Token (APIFY_TOKEN) with appropriate permissions

Setup steps:

1. Clone the repository (or install the npm package if you’re using the published version): git clone https://github.com/apify/mcp-server-rag-web-browser.git cd mcp-server-rag-web-browser

2. Install dependencies: npm install

3. Build the project (if required by the repository): npm run build

4. Run the MCP server locally (example): npx @apify/mcp-server-rag-web-browser

5. Set up environment variable for authentication (example shell): export APIFY_TOKEN=your-apify-api-token

6. Validate the server is reachable via MCP clients by providing the configured mcpServers entry (see mcp_config in this document).

Tips and notes:

• The server is designed to run in Standby mode with MCP clients; it expects standard MCP messaging on stdio.
• Ensure APIFY_TOKEN is set in your environment to access Apify APIs.
• If you plan to deploy or test against mcp.apify.com, you can migrate to the hosted service as documented in the MCP migration guide.
• For debugging, consider using the MCP Inspector tool mentioned in the repository, and run the server with a debug-enabled command if needed.
• If you encounter issues with Google search quota or page scraping, adjust the search arguments or try different scrapingTool settings as described in the Tools section of the README.