comfyui_mcp_server

This MCP server provides two core capabilities for ComfyUI: image generation using the ComfyUI text-to-image workflows and prompt optimization to improve generation results. It exposes an MCP-compatible endpoint that leverages the FastMCP framework to communicate with a locally running ComfyUI instance. Users can request new images by supplying textual prompts (and optional parameters like width, height, and seed), or request optimized prompts that can yield higher-quality outputs for a given concept. The server returns both local file paths and online-accessible URLs for the generated images, along with any relevant metadata such as seed and dimensions.

Prerequisites:

• Python 3.7+ installed
• ComfyUI instance running and accessible at the configured COMFY_URL
• FastMCP library installed in your Python environment

1. Clone the MCP server repository:

git clone https://github.com/yourorg/comfyui_mcp_server.git
cd comfyui_mcp_server

2. Create and activate a Python virtual environment (optional but recommended):

python -m venv venv
# Windows
venv\Scripts\activate.bat
# macOS/Linux
source venv/bin/activate

3. Install dependencies:

pip install -r requirements.txt

4. Configure the server (example shown in README):

# Edit the configuration file if needed
# The example uses:
# COMFY_URL=http://127.0.0.1:8188/

5. Run the server:

python D:\code\comfyui_mcp_server\server.py

6. Verify the MCP endpoint is reachable using your MCP client and start issuing generate_image or optimize_image_prompt requests.

Note: If you prefer a different runtime (e.g., a Docker-based setup), adjust the mcp_config accordingly and ensure COMFY_URL points to your running ComfyUI instance.

Tips:

• Ensure ComfyUI is reachable from the MCP server via the COMFY_URL you provide; firewall rules may block access.
• The server can automatically adjust image dimensions to multiples of 8; consider specifying width/height if you have strict requirements.
• Seed handling: a random seed is generated if not provided, which helps with output diversity.
• If you encounter timeout errors, you may increase the timeout in the mcp_config to accommodate slower image generation times.
• The server returns both local file paths and online URLs; you can store results or display images in a UI accordingly.