LEAP MCP is an AI-powered educational video generator that creates short, explainer-style videos on any topic with narration and 3Blue1Brown-inspired animations. It leverages OpenAI for scripting and Claude Code for interactive planning when used via Claude Code, and it renders visuals using Manim. To get started, clone the repository, install the Python package, and configure your API key. Once running, you can trigger video creation by providing natural language prompts (for example, 'Create an educational video about black holes'), and LEAP will generate a structured storyboard, narration, and corresponding animations. The workflow is designed to be used from within Claude Code or any MCP client that supports Python-based servers; use the client to invoke the /mcp command and select leap-mcp to connect, then issue prompts to generate topic-specific videos.

Prerequisites:

• Python 3.8–3.11
• A valid OpenAI API key
• Claude Code (for interactive planning and connection; optional if using MCP client directly)

Install and run:

1. Clone the repository: git clone https://github.com/sid-thephysicskid/leap-mcp cd leap-mcp

2. Install the package (as indicated by the README): python setup.py

3. Configure API keys:

   • If you skip during setup, copy the example env file and edit it: cp .env.example .env

     Edit .env to include your OpenAI API key:

     OPENAI_API_KEY=your_actual_key_here

4. Run the server via the Python setup command (as shown in the Quick Start): python setup.py

5. If using Claude Code for control:

   • Start Claude Code
   • In Claude Code chat, type /mcp and select leap-mcp to connect
   • If not connected, restart Claude Code and try /mcp again

Tips and notes:

• The project uses Python + Manim to render 3Blue1Brown-style animations. Ensure your Python environment supports Manim and that you’re using a compatible Python version (3.8–3.11).
• If you encounter API errors, double-check your OPENAI_API_KEY in the .env file and ensure it has the necessary permissions.
• The Voices available include nova (default), alloy, echo, fable, shimmer, and onyx; you can select different narrators for your videos.
• Troubleshooting steps: for connection issues, restart Claude Code and re-invoke /mcp; for Python version problems, revert to Python 3.8–3.11.
• This MCP server integrates with Manim for visuals and OpenAI for narration; ensure your environment has the required dependencies installed as described in setup.py.
• If you plan to extend support to additional MCP clients, keep an eye on future releases for Cursor/Windsurf/Qwen integration.