ComfyUI Launcher: Run Any Workflow with Zero Setup

TL;DR

TL;DR: ComfyUI Launcher is an open-source app that lets you run any ComfyUI workflow with zero setup — it auto-installs missing nodes, models, and keeps each workflow isolated.

Source and Accuracy Notes

All links verified from actual source. Do not trust paraphrased claims — always read the README yourself.

• Project page: comfyworkflows.com
• Source repository: github.com/ComfyWorkflows/ComfyUI-Launcher
• License: AGPL-3.0 (verified via GitHub API license.spdx_id)
• HN launch thread: news.ycombinator.com/item?id=38546467
• Source last checked: 2026-07-10 (README, GitHub API)

What Is ComfyUI Launcher?

ComfyUI is a powerful, node-based UI for running diffusion AI models. The workflow JSON files ComfyUI exports are self-contained — except when they depend on custom nodes or models you do not have installed. That is where ComfyUI Launcher helps.

ComfyUI Launcher is an open-source tool from ComfyWorkflows that runs any exported ComfyUI workflow with zero manual setup. It automatically installs missing custom nodes and model files. Each workflow runs in its own isolated environment, so updating one workflow’s dependencies cannot break another’s.

From the README:

Run any ComfyUI workflow w/ ZERO setup.

Setup Workflow

Option 1: Docker (Linux and Windows)

Requires Docker with GPU support.

docker run \
--gpus all \
--rm \
--name comfyui_launcher \
-p 4000-4100:4000-4100 \
-v $(pwd)/comfyui_launcher_models:/app/server/models \
-v $(pwd)/comfyui_launcher_projects:/app/server/projects \
-it thecooltechguy/comfyui_launcher

Then open http://localhost:4000 in your browser.

To update to a new version:

docker pull thecooltechguy/comfyui_launcher

Option 2: Manual (macOS, Linux, Windows WSL)

One-time installation:

git clone https://github.com/ComfyWorkflows/comfyui-launcher
cd comfyui-launcher/
./run.sh

Open http://localhost:4000 in your browser. To update later, just git pull.

Using a Reverse Proxy

If you need to expose the Launcher on a single port (for example, behind a reverse proxy), run with PROXY_MODE=true:

docker run \
--gpus all \
--rm \
--name comfyui_launcher \
-p 4000:80 \
-v $(pwd)/comfyui_launcher_models:/app/server/models \
-v $(pwd)/comfyui_launcher_projects:/app/server/projects \
-e PROXY_MODE=true \
-it thecooltechguy/comfyui_launcher

This routes all workflow projects through a single port using the built-in NGINX configuration.

How Workflow Sharing Works

When you export a workflow from ComfyUI Launcher, it saves a .json file. Anyone who receives that file can open it in their own ComfyUI Launcher installation and run it immediately — the Launcher handles all dependency installation automatically. This makes sharing complex AI image generation pipelines as easy as sharing a single JSON file.

Practical Evaluation Checklist

• Auto-installs custom nodes and models — confirmed from README Features section
• Zero-setup for imported workflows — confirmed: workflows exported by the tool require no manual setup from recipients
• Isolated environments per workflow — confirmed: each workflow runs in its own isolated environment
• Docker with GPU pass-through — confirmed: --gpus all flag in Docker run command
• Huggingface Spaces deployment available — confirmed: multimodalart/comfyUI-laucher-v2 on Hugging Face Spaces
• RunPod cloud deployment — confirmed: separate cloud/RUNPOD.md in repository
• Cross-platform — Linux, macOS, Windows (WSL) supported

Security Notes

• Workflow JSON files are code — only import workflows from sources you trust. A malicious workflow could execute arbitrary commands during node installation.
• The Docker container runs as the app user inside the container, but model/node files are mounted from your host filesystem — review the Dockerfile if you need stricter isolation.
• PROXY_MODE exposes port 80 inside the container to the host network — use firewall rules or a reverse proxy with authentication if exposing publicly.

FAQ

Q: Does this work without a GPU? A: Yes, but image generation will be extremely slow. The Docker command includes --gpus all for GPU acceleration. Without a CUDA GPU, you must use CPU inference, which is impractical for diffusion models.

Q: Can I use this on macOS without Docker? A: Yes. The manual setup option (Python 3) works on macOS natively without Docker. You just need to run ./run.sh after cloning the repository.

Q: What happens when a workflow’s custom node updates and breaks compatibility? A: Each workflow runs in its own isolated environment, so a node update affects only that specific workflow’s environment. Your other workflows continue to work as before.

Q: Is the project actively maintained? A: The GitHub repository shows its last commit in August 2024. The HN launch was in March 2024. Check the repo for recent activity before relying on it for production use.

Conclusion

ComfyUI Launcher solves the biggest friction point with ComfyUI workflows: dependency hell. Instead of manually installing custom nodes and hunting for model weights, the Launcher handles it automatically. The isolated environment model is particularly smart — it means you can collect workflows from different sources without worrying that upgrading one node will break another.

If you work with ComfyUI frequently or share workflows with a team, this is a significant quality-of-life improvement. The Docker path is the most reproducible; the manual Python path is easiest for macOS users without Docker.

Try it: github.com/ComfyWorkflows/ComfyUI-Launcher