Pixelle MCP is an AIGC solution built around the MCP protocol, delivering a unified, omnimodal agent framework that converts ComfyUI workflows into MCP tools without writing code. It supports both local ComfyUI deployments and cloud-based RunningHub workflows, enabling users to execute and manage workflows regardless of hardware or hosting. The system integrates a Chainlit-based web interface for interactive multimodal sessions and exposes an MCP endpoint so external MCP clients (such as Cursor and Claude Desktop) can connect and operate tools created from workflows. The architecture combines a web UI, an MCP server, and a file service, plus a workflow engine that automatically translates ComfyUI canvases into executable MCP tools. It emphasizes zero-config deployment via PyPI, CLI, and Docker, with environment-variable configuration. It also supports multiple LLM providers (OpenAI, Ollama, Gemini, DeepSeek, Claude, Qwen, and more), and a flexible deployment flow with both one-click experiences and a full local development path. The project aims to enable full TISV multimodal generation and scalable operation through cloud and local modes.