Comfy UI Builder

You can run and integrate the MCP server that adds seed knowledge, dynamic workflows, real-time execution, and a broad set of tools for ComfyUI and similar interfaces. This MCP server enables node discovery, template-based workflows, model management, and WebSocket-driven execution with efficient network usage, making it easier to build and run complex pipelines.

How to use

You will start by building the project to generate the MCP server artifacts, then run the server and connect it to your MCP client or CURSOR/Claude workflow. Once running, you will discover nodes, assemble workflows, and execute them with real-time progress updates.

How to install

Prerequisites you need before installing: Node.js 18 or newer.

Install the package from npm and set up the server:

npm install mcp-comfy-ui-builder

# Or install from source
git clone https://github.com/MIt9/mcp-comfy-ui-builder.git
cd mcp-comfy-ui-builder
npm install

Running the MCP server

Build the project first so the knowledge base and tooling are ready for use.

npm run build

Then start the MCP server. You have two common options shown in the project flow.

npm run mcp

# Or, without npm
node dist/mcp-server.js

Connect to the MCP server

Configure your MCP client to connect to the local server. If you are using a local ComfyUI instance, you can set the COMFYUI_HOST environment variable to point to it.

export COMFYUI_HOST="http://localhost:8188"

Publish to the MCP Registry

If you plan to publish your MCP server, ensure you provide a server.json configuration and an mcpName as part of your MCP Registry integration.

Available tools

list_node_types

List available node types supported by the MCP server.

get_node_info

Retrieve detailed information about a specific node.

check_compatibility

Check compatibility of nodes with the current environment.

suggest_nodes

Suggest suitable nodes based on a given context or input.

discover_nodes_live

Discover currently running or available nodes in live mode.

search_nodes

Search the knowledge base for nodes matching criteria.

sync_nodes_to_knowledge

Sync the currently discovered nodes into the knowledge base.

create_workflow

Create a new dynamic workflow.

add_node

Add a node to an existing workflow.

connect_nodes

Connect the output of one node to the input of another within a workflow.

validate_workflow

Validate a workflow for correctness before execution.

list_templates

List pre-built templates available for workflows.

build_workflow

Build a workflow from a template and parameters.

create_template

Create a parameterized template for reuse.

apply_template

Apply a template to a workflow with specific parameters.

list_macros

List reusable sub-workflows (macros).

insert_macro

Insert a macro into a workflow.

execute_workflow_sync

Execute a workflow with WebSocket progress streaming.

execute_workflow_stream

Execute a workflow and stream full event history.

get_execution_progress

Track real-time progress with node-level granularity.

execute_batch

Run multiple workflows concurrently with reduced network traffic.

execute_chain

Run sequential workflows with data passing between steps.

get_system_resources

Query GPU/VRAM/RAM resources and recommendations.

list_models

List available models, including checkpoints, LoRA, VAEs, and more.

check_model_exists

Check if a specific model exists in the system.

get_workflow_models

Get models associated with a particular workflow.

list_outputs

List available outputs from executions.

download_output

Download a single output file.

download_all_outputs

Download all outputs from an execution.

list_queue

View the execution queue.

interrupt_execution

Interrupt a running execution.

clear_queue

Clear the execution queue.

list_plugins

List available data-only plugins.

reload_plugins

Reload the plugin set.

install_custom_node

Install a custom node via ComfyUI-Manager.

install_model

Install a model via ComfyUI-Manager.