Comfy UI MCP Server

The Jonpojonpo Comfy UI MCP Server bridges local ComfyUI installations with AI assistants by exposing a lightweight, note‑centric data store over the Model Context Protocol. By turning each note into a resource, the server turns an otherwise opaque UI into a structured API that Claude and other MCP‑enabled assistants can interrogate, update, and summarize. This solves a common pain point for developers who want to keep contextual information—such as experiment logs, model settings, or research snippets—synced across tools without writing custom integrations.

At its core the server offers three simple yet powerful capabilities. First, it implements a note storage resource: each note has a unique name, description, and plain‑text content. The custom URI scheme lets assistants retrieve or modify any note directly with a standard or . Second, the summarize‑notes prompt aggregates all stored notes into a concise or detailed summary based on the optional argument. This allows assistants to quickly generate overviews of a project’s current state or history, saving time on manual documentation. Third, the add‑note tool lets users append new notes by supplying a name and content; the server updates its state and broadcasts changes to any connected clients, ensuring that all assistants stay in sync.

Developers can use the server in a variety of real‑world scenarios. In research pipelines, an assistant could automatically log experiment parameters as notes and later pull a summary for reporting. In creative workflows, designers might capture mood board ideas or iteration notes and let the assistant retrieve them during brainstorming sessions. Because the server communicates purely over stdio, it can run on any platform that supports MCP without requiring network configuration, making it ideal for local or isolated environments.

Integration with existing AI workflows is straightforward. An MCP‑compatible client, such as Claude Desktop or any custom assistant built on the protocol, can register the server’s URI and immediately access the note resources. The prompt and tool expose a clean API surface that can be invoked programmatically or through conversational commands, enabling seamless chaining of data retrieval, manipulation, and summarization within a single dialogue. The server’s notification mechanism further ensures that assistants react to changes in real time, maintaining a consistent state across multiple tools.

What sets this MCP server apart is its focus on structured knowledge management within a local UI ecosystem. By turning notes into first‑class resources and providing automated summarization, it eliminates the need for external databases or manual documentation. The result is a lightweight, extensible bridge that lets developers and AI assistants collaborate on the same information space with minimal friction.