Deep Knowledge MCP Server Give your AI the power to browse, read, and understand technical documentation and code repositories. This Apify Actor implements the Model Context Protocol (MCP), allowing LLMs (like Claude) to perform deep research on technical topics by leveraging Apify's powerful scraping infrastructure. --- ## 🚀 Features * 🔍 Search Technical Docs: Finds relevant documentation, libraries, and GitHub repositories using a specialized Google Search setup. * 📖 Fetch Documentation: Extracts clean, markdown-formatted content from any documentation URL, perfect for LLM context windows. * ⚡ Fast & Scalable: built on Apify's infrastructure, handling anti-scraping protections automatically. * 💰 Pay-per-Event: Simple pricing model based on the tools you use. --- ## 🛠️ Tools Provided ### 1. search_technical_docs Searches for technical resources. * Input: query (e.g., "LangChain python documentation"), max_results (default: 5). * Output: A list of titles, URLs, and descriptions. ### 2. fetch_documentation Reads a web page and returns its content. * Input: url (e.g., "https://python.langchain.com/docs/get_started/introduction"). * Output: The page content converted to Markdown. --- ## 💻 How to Use ### Option 1: limit with Claude Desktop (Recommended) To use this Actor as a tool in Claude Desktop: 1. Deploy this Actor to your Apify account. 2. Get your Apify Token from Settings > Integrations. 3. Configure your claude_desktop_config.json: json { "mcpServers": { "deep-knowledge": { "command": "npx", "args": [ "-y", "@apify/mcp-server-runner", "ACTOR_NAME_OR_ID", "--token", "YOUR_APIFY_TOKEN" ] } } } (Replace ACTOR_NAME_OR_ID with this Actor's name, e.g., eager_cornet/deep-knowledge-mcp-server) ### Option 2: Standalone Mode (Demo) You can run this Actor directly on the Apify Platform to test it without an AI client. 1. Go to the Input tab. 2. Uncheck Run as MCP Server. 3. Enter a Demo Search Query. 4. Click Start. 5. View results in the Output tab. --- ## 💸 Monetization & Pricing This Actor uses Pay-per-Event pricing. You are charged only when the AI successfully calls a tool. * Search: charged per successful search operation. * Fetch: charged per successful page fetch. Check the Pricing tab for exact rates. --- ## 👨‍💻 Development This server is built with Python using the mcp SDK and apify-client. Local Development: bash # Install dependencies pip install -r requirements.txt # Run locally (requires APIFY_TOKEN env var) python src/main.py