youtube-intelligence-mcp

# YouTube Intelligence MCP An autonomous Model Context Protocol (MCP) server that transforms YouTube tutorials into executable workflows and fully replicated projects. It uses a multi-agent system (Playwright, ChatGPT, and Gemini) to extract, analyze, and execute tutorial steps. ## Features - **Automated Extraction**: Pulls transcripts, metadata, links, and captures screenshots from YouTube videos. - **Deep Analysis**: Uses Gemini Pro Vision (via Web UI) to analyze video frames and transcript context. - **Autonomous Execution**: ChatGPT Orchestrator plans and executes terminal commands and file operations. - **Self-Healing**: Robust browser management with CDP connection and automatic page recovery. --- ## Prerequisites Before you begin, ensure you have the following installed: - **Node.js**: v18.0.0 or higher - **npm**: v9.0.0 or higher - **Google Chrome**: A recent version (needed for CDP connection) - **Playwright Browsers**: Even though it connects via CDP, local drivers are required. --- ## Installation Follow these steps to set up the YouTube Intelligence MCP server on your machine: ### 1. Clone the Repository git clone https://github.com/sarathi-eng/youtube-intelligence-mcp.git cd youtube-intelligence-mcp ### 2. Install Dependencies npm install ### 3. Build the Project ```bash npm run build ``` --- ## Configuration ### MCP Client Setup (e.g., Claude Desktop) #### Option A: Using `npx` (Recommended for latest version) Add the following to your `claude_desktop_config.json`: ```json { "mcpServers": { "youtube-intelligence": { "command": "npx", "args": ["-y", "youtube-intelligence-mcp"], "env": { "NODE_ENV": "production" } } } } ``` #### Option B: Local Path Add the following to your `claude_desktop_config.json`: ```json { "mcpServers": { "youtube-intelligence": { "command": "node", "args": ["/absolute/path/to/youtube-intelligence-mcp/build/index.js"], "env": { "NODE_ENV": "production" } } } } ``` > **Note**: Replace `/absolute/path/to/` with the actual path on your system. --- ## Usage Once configured, the following tool becomes available to your MCP-enabled AI: ### `replicate_tutorial` **Arguments**: - `url` (string): The full URL of the YouTube tutorial you want to replicate. **Example**: "Replicate this tutorial for me: https://www.youtube.com/watch?v=dQw4w9WgXcQ" ### What Happens Next? 1. **Extraction**: The server opens a persistent Chrome instance and extracts all data. 2. **Analysis**: Data is sent to Gemini to generate a structured execution graph. 3. **Execution**: ChatGPT receives the graph and starts executing steps (creating files, running bash commands). 4. **Final Output**: A full report is saved to `replication_results/replication_report.md`. --- ## Technical Architecture - **Runtime Agent**: Manages the Playwright/Chrome lifecycle. - **ChatGPT Agent**: Interacts with the ChatGPT Web UI for high-level reasoning and command generation. - **Gemini Agent**: Interacts with the Gemini Web UI for vision-based video analysis. - **Execution Engine**: The central "brain" that coordinates the 7-phase workflow. ---