ComfyUI MCP serves as a local-first, agent-native control plane designed for ComfyUI, providing an MCP server and a Claude Code plugin. Its primary function is to enable users to generate images, video, and audio; author and execute workflows; and manage models and custom nodes through natural language commands. The platform is intended for users who wish to control ComfyUI instances, whether running locally, remotely, or via Comfy Cloud, and it facilitates interaction with AI assistants to operate ComfyUI beyond simple prompt-to-image tasks.
Key features include the ability to generate text-to-image outputs with support for ControlNet and IP-Adapter conditioning, author workflows using templates or a compact DSL, modify and validate workflows, and visualize or round-trip them. Users can manage models by searching repositories such as HuggingFace and CivitAI, downloading, listing, or removing models as needed. The tool also supports installing, updating, snapshotting, and bisecting custom node packs via ComfyUI-Manager. Additionally, it provides capabilities for installing and updating ComfyUI itself, inspecting the environment, and starting or stopping the server.
ComfyUI MCP can be deployed across local, remote, or cloud environments. It automatically detects local installations on macOS, Linux, and Windows, and can connect to remote instances using configuration options such as --comfyui-url or interact with Comfy Cloud through an API key. The setup process involves installing ComfyUI and configuring the MCP server, with support for environment variables, different transports, authentication tokens, and remote targeting. The tool offers a single configuration to target all three deployment modes, allowing users to select the setup that best matches their workflow requirements.
The project is built and maintained by @artokun, a contributor to the ComfyUI ecosystem. The source code and issue tracking are available on GitHub. No information regarding pricing or licensing is provided in the available documentation.
In the Other AI space, ComfyUI MCP takes a focused approach. It focuses on automating and managing ComfyUI workflows and models using natural language and agent-native controls. It is built as an open-source project for AI developers and workflow engineers. ComfyUI MCP is open source under the MIT license. ComfyUI MCP is available on the command line and API, and it can be self-hosted.
artokun builds and maintains ComfyUI MCP, and the product first shipped in 2026. Development happens publicly on GitHub with 235 stars and 396 commits in the last 90 days. PulseGate's similarity index finds few close equivalents — ComfyUI MCP occupies a relatively distinct niche. Key capabilities include workflow automation, model management, and natural language control. It exposes integrations via an MCP server.
Latest indexed changes and source events
artokun/comfyui-mcp discovered by the PulseGate indexer
Other apps tracked under the same category.
