Pixelle

About

Pixelle is an open-source, omnimodal agent framework that seamlessly integrates ComfyUI with Large Language Models (LLMs) via the Model Context Protocol (MCP). It allows users to convert complex ComfyUI workflows into callable MCP tools with zero code, empowering LLMs to perform a wide range of AIGC tasks across text, image, sound/speech, and video. Built on the extensible ComfyUI ecosystem, Pixelle provides a flexible and unified solution for developing, deploying, and utilizing multimodal AI generation capabilities through a powerful client-server architecture.

Key Features

• Supports full-modal (Text, Image, Sound/Speech, Video) conversion and generation.
• Built on ComfyUI, inheriting its entire open ecosystem capabilities.
• Enables zero-code development and dynamic addition of new MCP tools from ComfyUI workflows.
• Provides a robust MCP Server for integration with any MCP client (e.g., Cursor, Claude Desktop).
• Offers flexible deployment options, including standalone server, standalone client, or combined.
• 95 GitHub stars

Use Cases

• Integrating advanced AI generation capabilities into existing MCP-compatible applications and platforms.
• Transforming custom ComfyUI workflows into easily callable AI tools for LLMs.
• Building multimodal AI agents capable of generating and converting various content types.