comfyui_LLM_party

MCP.Pizza Chef: heshengtao

Built as an add-on for ComfyUI, the drag-and-drop canvas normally used for making images, this pack adds blocks for language models: chat, role-playing personas, searchable knowledge bases from your own files, spoken audio, image reading, and links to QQ, Feishu, and Discord. It can also borrow tools from MCP add-ons you already run. Hosted models such as GPT, Gemini, Qwen, and DeepSeek need your own key; models running on your own hardware do not.

Communication
Images/Design
Writing

Use This MCP client To

Build a chat assistant that answers from my own documents Give a character I designed a speaking voice Read an image and write a matching prompt automatically Run a language model on my own computer, offline Connect an assistant to my Discord or Feishu group Chain several assistants together to finish one job

README

图片

C‌‌​‎​‎‏​‍‎​‎​‎‏​‌‎​‎‍​‍‏​‍‌​‌‏omfyui_llm_party aims to develop a complete set of nodes for LLM workflow construction based on comfyui as the front end. It allows users to quickly and conveniently build their own LLM workflows and easily integrate them into their existing image workflows.

Effect display

comfyui_LLM_party FAQ

Do I need ComfyUI installed first?
Yes, this is an add-on for ComfyUI, though a Windows bundle exists that includes both.
Do I need a paid key?
Only for hosted models like GPT or Gemini. Models running on your own computer need no key.
Can I use this to answer questions from my own documents?
Yes — it includes knowledge-base blocks that search your files before the model replies.
Is this an MCP server or a client?
A client. It picks up tools from MCP add-ons you already run and hands them to the model.
How hard is the setup?
Expect a real install: Python packages, model files, and a careful read of the project page.
Does it work with models running offline?
Yes, including local text models and local image-reading models.
Can it talk to chat apps?
It has blocks for QQ, Feishu, and Discord.