ComfyUI MCP Bridge

Example Use Cases

• 1

  Generate images via a local ComfyUI instance and receive the resulting image URL for downstream use.

• 2

  Fetch generated image PNG binaries and obtain textual captions using a local ollama vision model for content understanding.

• 3

  Extract WD1.4-style tags from images (via wd-tagger) to support tagging, filtering, or iterative prompt refinement workflows.

Description

This project implements an MCP server to interact with ComfyUI's API, providing image generation, PNG retrieval, captioning (via ollama vision models), and WD1.4 tag extraction. It exposes tools like generate_picture, get_picture, get_caption, and get_tag, supports custom ComfyUI workflows, and configurable model/Lora/sampling settings. Designed for local setups (requires local ComfyUI, ollama, and uv), it enables agents to incorporate image generation and analysis into automated workflows while keeping models and data local.