Drawio MCP

A modular client-server setup for generating Draw.io diagrams using natural language prompts. It uses LangGraph, LangChain, Gemini 2.5, and a FastMCP-based server to convert text into XML diagram code and serve it via a tool-based API.

---

🔧 Features

• Streamlit client for chatting with the tool.
• Server powered by FastMCP that:
  • Parses user prompt → instructions
  • Generates draw.io XML from instructions
  • Verifies XML
• Uses gemini-2.5-flash-preview-05-20 for generation.
• Output is saved as .drawio inyour ~/Downloads folder.

---

🗂 Structure

├── client.py # Streamlit UI

├── server.py # FastMCP-based server

├── Dockerfile # Containerization setup

├── requirements.txt # Python dependencies

├── prompts/ # Prompt templates

├── LICENSE

├── README.md

---

🚀 Running Locally

1. Install Python deps

pip install -r requirements.txt

2. Set environment variables

Make sure .env includes:

1. GROQ_API_KEY=your_key_here

2. GOOGLE_API_KEY=your_key_here

3. Start the server

python server.py

4. Run the client

python client.py

5. Docker Support

docker build -t drawio_mcp .
docker run -p 8000:8000 --env-file .env drawio_mcp

6. ✍🏾 Prompt Example

Create a workflow showing user login, verification, and dashboard redirection.