Skip to content

Quickstart

Get up and running with ComfyGit in 5 minutes. By the end, you'll have created an environment, added custom nodes, and understand the basics of version control for your ComfyUI workflows.

Before you begin

Make sure you have:

  • ComfyGit installed — Installation guide
  • A terminal or command prompt open
  • Internet connection for downloading dependencies

Step 1: Initialize your workspace

Create a ComfyGit workspace:

cg init

This creates ~/comfygit/ with the following structure:

~/comfygit/
├── comfygit_cache/        # Cache for registry, custom nodes, comfyui clones, etc.
├── models/                # Shared models directory
├── environments/          # Your ComfyUI environments
├── logs/                  # Environment and workspace logs
└── .metadata/             # Workspace configuration

Custom workspace location

Use a different path: cg init /path/to/workspace

Step 2: Create your first environment

Create an isolated ComfyUI environment:

cg create my-project --use

This will:

  1. Create a new environment called my-project
  2. Download and install ComfyUI (latest release from github)
  3. Install PyTorch with GPU support (auto-detected)
  4. Set it as your active environment
  5. Take 2-5 minutes depending on your internet speed

What's happening?

ComfyGit is creating an isolated Python environment with UV, downloading ComfyUI from GitHub, and installing dependencies. The --use flag makes this your active environment.

What you'll see:

🚀 Creating environment: my-project
   This will download PyTorch and dependencies (may take a few minutes)...

✓ Environment created: my-project
✓ Active environment set to: my-project

Next steps:
  • Run ComfyUI: cg run
  • Add nodes: cg node add <node-name>

Step 3: Run ComfyUI

Start ComfyUI in your environment:

cg run

ComfyUI opens at http://localhost:8188

Run in background

cg run &

Or use screen/tmux to keep it running: 

screen -S comfy
cg run
# Detach with Ctrl+A, D

Step 4: Check environment status

Open a new terminal and check your environment's status:

cg status

Output:

Environment: my-project ✓

✓ No workflows
✓ No uncommitted changes

Step 5: Add custom nodes

Let's add a custom node from the ComfyUI registry:

cg node add comfyui-depthflow-nodes

This will:

  1. Look up the node in the ComfyUI registry
  2. Clone the repository to custom_nodes/
  3. Install Python dependencies
  4. Update pyproject.toml

Adding nodes from GitHub

# Add by GitHub URL
cg node add https://github.com/akatz-ai/ComfyUI-AKatz-Nodes

# Add specific version/branch
cg node add comfyui-depthflow-nodes@main

Avoid ComfyUI-Manager

ComfyGit replaces ComfyUI-Manager's functionality. Don't install comfyui-manager - use cg node add instead.

Try adding more nodes:

# Popular nodes to try
cg node add comfyui-impact-pack
cg node add comfyui-controlnet-aux

Step 6: Commit your changes

Save your environment's current state:

cg commit -m "Added depthflow nodes"

This creates a git commit in the .cec/ directory tracking:

  • Custom nodes and their versions
  • Python dependencies
  • Model references
  • Workflow files

Check your commit history:

cg log

Output:

Version history for environment 'my-project':

v2: Added depthflow nodes
v1: Initial environment setup

Use 'cg rollback <version>' to restore to a specific version

Verbose mode

cg log --verbose

Shows timestamps and full commit hashes.

Step 7: Experiment safely

Let's add another node and see how rollback works:

# Add a test node
cg node add comfyui-video-helper-suite

# Check status
cg status

Now roll back to remove that change:

cg rollback v1

Your environment reverts to the state from your first commit—comfyui-video-helper-suite is removed automatically.

Discard uncommitted changes

# Discard all changes since last commit
cg rollback

Step 8: Load a workflow

Let's resolve dependencies for a workflow. Download a sample workflow JSON file, then:

# Move workflow to ComfyUI/user/default/workflows/
cp /path/to/workflow.json ~/comfygit/environments/my-project/ComfyUI/user/default/workflows/

# Resolve dependencies
cg workflow resolve workflow.json

ComfyGit will:

  1. Analyze the workflow JSON
  2. Identify required nodes
  3. Prompt you to install missing nodes
  4. Find required models
  5. Suggest download sources

Auto-install mode

cg workflow resolve workflow.json --install

Automatically installs all missing nodes without prompting.

Common workflows

Now that you have the basics, here are some common tasks:

Switch between environments

# Create another environment
cg create testing --use

# List all environments
cg list

# Switch back to my-project
cg use my-project

Update a custom node

# Update to latest version
cg node update comfyui-depthflow-nodes

# View installed nodes
cg node list

Add Python dependencies

# Add a package
cg py add requests

# Add from requirements.txt
cg py add -r requirements.txt

# List installed packages
cg py list

Export your environment

Share your environment with others:

cg export my-workflow-pack.tar.gz

This creates a tarball containing:

  • Node metadata and versions
  • Model download URLs
  • Python dependencies
  • Workflow files

Import on another machine:

cg import my-workflow-pack.tar.gz --name imported-env

Essential commands

Here are the most important commands for daily use:

Command What it does Example
cg create Create new environment cg create prod --use
cg use Switch active environment cg use testing
cg list List all environments cg list
cg run Start ComfyUI cg run
cg status Show environment status cg status
cg node add Add custom node cg node add comfyui-depthflow-nodes
cg commit Save current state cg commit -m "message"
cg log List environment commits cg log
cg rollback Revert to previous state cg rollback v1
cg export Export environment cg export my-pack.tar.gz
cg import Import environment cg import my-pack.tar.gz

See the CLI reference for a complete list of commands.

Pro tips for beginners

Use tab completion

Install shell completion for faster typing: 

cg completion install

Then use Tab to autocomplete environment names, node names, and commands.

Check logs when things fail

cg debug -n 50

Shows the last 50 log lines for debugging.

Start with a clean environment

Don't add too many nodes at once. Start minimal, add what you need, commit often.

Use descriptive commit messages

cg commit -m "Added IPAdapter for style transfer"

Makes it easy to find specific versions later.

Specify PyTorch backend manually

# For NVIDIA GPUs
cg create gpu-env --torch-backend cu128

# For AMD GPUs
cg create amd-env --torch-backend rocm6.3

# For CPU only
cg create cpu-env --torch-backend cpu

What's next?

Now that you've learned the basics, explore more advanced features:

Getting help

  • In your terminal: Run cg --help or cg <command> --help
  • Documentation: You're here! Browse other guides
  • Issues: Report bugs on GitHub Issues
  • Discussions: Ask questions on GitHub Discussions