Log inSubmit server
Pixelle MCP logo

Pixelle MCP

by AIDC-AI

Converts ComfyUI workflows into MCP tools for text, image, sound, and video generation through local or RunningHub execution.

InstallRepositoryWebsite
Full README Compare alternatives Embed widget Claim this listing

Install Pixelle MCP

Pick your client, then add it in one click or paste the config.

STDIOPython
~/Library/Application Support/Claude/claude_desktop_config.json
claude_desktop_config.json
{
  "mcpServers": {
    "aidc-ai-pixelle-mcp": {
      "command": "uvx",
      "args": [
        "pixelle@latest"
      ],
      "env": {
        "PORT": "YOUR_VALUE_HERE"
      }
    }
  }
}

Settings → Developer → Edit Config, merge this into mcpServers, then fully quit and reopen Claude Desktop.

Values are written into the snippet above so you can paste it as is. They never leave your browser.

We couldn’t confirm it starts correctly

We ran the install command but it didn't respond within our test window — this can mean a slow first-time install rather than a real problem.

uvx pixelle@latest

No response to initialize.

This is an experimental automated check and can have false negatives — missing environment variables, a slow cold install, etc. It doesn’t necessarily mean something’s wrong. Last checked 25d ago.

Overview

AIDC-AI/Pixelle-MCP MCP server converts ComfyUI API workflows or RunningHub workflow IDs into MCP tools for multimodal generation. It can execute workflows through a local ComfyUI installation or the RunningHub cloud service, while a Chainlit interface provides web-based interaction and file handling. Reach for it when an agent needs configurable text, image, sound, or video workflows without writing a separate MCP tool for each one. It also exposes an MCP endpoint for clients such as Cursor and Claude Desktop.

Use cases

  • Convert ComfyUI image workflows into agent-callable MCP tools
  • Run multimodal generation through a local ComfyUI installation
  • Execute ComfyUI workflows through RunningHub without local GPU hardware
  • Expose custom text, sound, or video workflows to Cursor or Claude Desktop

Key features

  • Converts ComfyUI API workflows into MCP tools
  • Supports local ComfyUI and RunningHub execution
  • Handles text, image, sound, and video workflows
  • Provides a Chainlit web interface and MCP endpoint
  • Supports workflow-defined parameters and file uploads
  • Offers CLI, PyPI, and Docker deployment options

Security & health

What it can reach and how reliably it runs — from our own checks, not the maintainer’s claims.

ActiveLast checked Today

Recent health check succeeded.

29/29 checks healthy over the last 78d

Supply-chain signal

No high-severity advisories surfaced by our automated scan.

Critical 0High 0Medium 0Low 0

Scanned 9d ago via OSV.dev · pixelle@latest (PyPI)

Adoption & maintenance

Factual signals from GitHub, npm, and our automated checks — not a rating.

GitHub stars1.1kStargazers on the source repository.
Last commit10mo agoMost recent push to the default branch.
Install checkInconclusiveDidn't respond in our test window — often a slow first install.
Directory activity7 viewsConfig copies, upvotes, and views on AllMCPs.

How Pixelle MCP works

What AIDC-AI/Pixelle-MCP MCP server does

AIDC-AI/Pixelle-MCP MCP server turns ComfyUI workflows into callable MCP tools. The framework supports workflows for text, images, sound or speech, and video, provided the underlying ComfyUI workflow and its models support those operations. Each workflow can become a separate tool without implementing a new tool service in code.

The project combines an MCP endpoint, a Chainlit-based web interface, file upload and download handling, and a workflow execution layer in one application. Agents can use the generated tools through an MCP client, while users can interact with the same application through the web interface.

How it works

For local execution, design a workflow in ComfyUI and export it in API format. Node titles use Pixelle’s parameter syntax to identify tool inputs, required values, upload-related parameters, and descriptions. The application reads that workflow and generates an MCP tool from it. A workflow should be tested in ComfyUI before being added, because an invalid or non-working workflow causes execution to fail.

RunningHub provides a second execution path. Instead of supplying a local workflow file, you provide the corresponding RunningHub workflow ID. This allows workflow execution without a local GPU or ComfyUI installation, but requires a RunningHub account and API key according to the project’s setup flow.

AIDC-AI/Pixelle-MCP MCP server also supports multiple LLM providers through its LiteLLM-based configuration. The README names OpenAI, Ollama, Gemini, DeepSeek, Claude, and Qwen among the supported providers. At least one LLM provider must be configured during initial setup.

Setup and configuration

The project supports several installation paths:

  • Run uvx pixelle@latest for a temporary CLI-based start.
  • Install the package with pip install -U pixelle, then run pixelle.
  • Clone the repository and start it with uv run pixelle for local development.
  • Copy .env.example to .env and use Docker Compose for container deployment.

The startup configuration wizard guides selection of the execution engine and LLM settings. The default web interface is available at http://localhost:9004, and the MCP endpoint is exposed at http://localhost:9004/pixelle/mcp. The port can be changed with the PORT environment variable. The default web login is dev for both username and password and can be changed after startup.

Tools and capabilities

AIDC-AI/Pixelle-MCP MCP server does not present a fixed, independently documented tool catalog. Its tools are generated from the workflows you add. Supported workflow inputs can include required parameters, descriptions, and image or other file uploads through the node-title DSL. The resulting tools can be used by MCP clients, including Cursor and Claude Desktop.

The application supports custom local workflows and RunningHub workflow IDs. Its Chainlit interface also supports multimodal interaction and integration with additional MCP servers, as stated in the project documentation.

Limitations and notes

Local mode requires a working ComfyUI environment and suitable models or nodes for the workflows you intend to run. RunningHub mode avoids local GPU and ComfyUI setup but depends on the cloud service and its API access. Workflow files must use ComfyUI’s API format when added locally. The documentation does not provide a fixed list of generated tool names; available tools depend on the workflows configured in the application.

Read the full README View source on GitHub

Reviews

No reviews yet — be the first to share how it worked for you.

Frequently Asked Questions about Pixelle MCP

AIDC-AI/Pixelle-MCP is an MCP server that connects AI clients to local ComfyUI or RunningHub cloud workflows. Its main tools are workflow-generated MCP tools for text, image, sound or speech, and video tasks, with Chainlit web access and file handling provided by the unified application.

Install it with `pip install -U pixelle` and run `pixelle`, or start it temporarily with `uvx pixelle@latest`. For development, clone the repository and run `uv run pixelle`; Docker deployment uses `.env` and Docker Compose.

Yes, AIDC-AI/Pixelle-MCP provides an MCP endpoint for Cursor. After starting the application, connect the client to `http://localhost:9004/pixelle/mcp`, using a different host or port if your deployment changes the defaults.

Yes, use RunningHub cloud mode to execute workflows without a local GPU or ComfyUI installation. This mode requires a RunningHub account and API key, while local mode runs against a self-hosted ComfyUI environment.

Export the workflow from ComfyUI in API format, define inputs and outputs with the supported node-title syntax, and add the workflow through the application. With RunningHub, provide the workflow ID instead of uploading a local workflow file.

Related MCP servers

All in Multimedia Process All alternatives
Add a badge to your READMEShow it’s listed — live SVG, Markdown or HTML.

AllMCPs Directory Badge

Full Badge Customizer

Showcase your server listing on GitHub or your project documentation. Embed this dynamic SVG badge to highlight official listing status and live engagement.

Badge Style:
Live Dynamic SVG PreviewPixelle MCP AllMCPs Directory Badge
Markdown (GitHub README)
[![AllMCPs](https://allmcps.com/api/badge/aidc-ai-pixelle-mcp?style=directory)](https://allmcps.com/mcp/aidc-ai-pixelle-mcp)
HTML Embed
<a href="https://allmcps.com/mcp/aidc-ai-pixelle-mcp"><img src="https://allmcps.com/api/badge/aidc-ai-pixelle-mcp?style=directory" alt="Pixelle MCP on AllMCPs" /></a>