RunComfy
RunComfy · Image · Leaning
RunComfy runs ComfyUI workflows on cloud GPUs, so you can run ComfyUI online with no local setup. Upload a workflow JSON and it rebuilds every node and model on 16 to 141 GB hardware.
About RunComfy
What Is RunComfy
RunComfy is a hosted ComfyUI platform for people who want the node graph workflow without owning a capable GPU or babysitting a Python environment. You bring a workflow, it provisions a private machine with the models and custom nodes that workflow expects, and you press run. The ComfyUI underneath is unmodified, so a graph you build here opens locally and the reverse also holds.
The problem it solves is the setup tax. Getting ComfyUI to a working state means matching CUDA builds, chasing missing nodes, and downloading multi-gigabyte checkpoints before you see a single image. On a laptop or a Mac, some video models won't run at usable speed at all. RunComfy moves that work to its servers and keeps the result, so the next session starts where the last one ended.
What it doesn't fix is complexity. ComfyUI is a node editor, and RunComfy doesn't hide that. If you'd rather type a prompt and get an image, a simpler generator will be less work. The platform pays off when your workflow has ControlNet branches, upscaling passes, or a second model stitched in.
Running ComfyUI online is the whole offer, and workflow hosting is the part people underrate. A workflow you can't reproduce is a workflow you can't sell, teach, or build on. RunComfy treats the saved environment as the deliverable, not the screenshot.
Getting Started
- Sign in at runcomfy.com and open the ComfyUI cloud workspace. A shared balance covers GPU time, the model playground, and training, so you top up once.
- Choose a starting point. Pick from the template library, or drop your own workflow.json into the uploader and let the auto-setup agent resolve the dependencies.
- Pick a machine. Sizes run from a 16 GB card for everyday image work up to H200 class hardware at 141 GB for large video models.
- Run the graph and adjust nodes in the native ComfyUI interface, then use Cloud Save to freeze the whole environment as a reusable snapshot.
- Deploy or share. Turn the saved workflow into a serverless API endpoint with a deployment ID, or send the link to a teammate so they open your exact setup.
Product Information
A quick look at RunComfy's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- Creators without a strong GPU
- ComfyUI power users who rebuild environments often
- Developers shipping image or video features
Tasks
- Running a heavy video workflow
- Sharing a working setup with a class or a team
- Training a LoRA on rented hardware
- Prototyping an app around image generation
Scenarios
- A studio that needs burst capacity
- A course or workshop where every student must see the same output
- A solo creator testing a new model release
Key features
One Click Setup from a Workflow JSON
This is the feature that sets RunComfy apart. Upload a workflow file and the auto-setup agent reads the graph, works out which custom nodes and checkpoints it expects, and installs them on a fresh machine. No Dockerfile, no pip list to reconcile. The promise is that a workflow shared on a forum three months ago still opens and runs, which rarely happens on a hand-built local install.
Reproducible Saved Environments
Cloud Save bundles the workflow JSON with the operating system, the Python stack, the ComfyUI version, the custom nodes, and the models into one snapshot. Reopening it later restores that exact combination. The practical effect is that you can keep several conflicting setups alive at once, one with an older node pack and one with the current release, and switch between them without breaking either.
Native ComfyUI on Machines from 16 GB to 141 GB
The editor is the stock ComfyUI build, kept in step with official releases, so nothing has been reskinned or hidden. Hardware spans six machine types, from a T4 or A4000 with 16 GB of VRAM up to an H200 with 141 GB and 240 GB of system memory. You pick per session, which means a quick upscale job doesn't have to be billed at video-model rates.
ComfyUI Serverless API from Any Saved Workflow
A saved workflow can be published as a ComfyUI serverless API with a unique deployment ID. RunComfy handles versioning, rolling updates, and autoscaling down to zero when traffic stops. Calls go to an endpoint such as api.runcomfy.net, with overrides that swap prompt text or input images per request. For a product team, that's the difference between a demo and something that survives a traffic spike.
Model Downloads and a Per-Generation Playground
Models pull directly from Civitai, Hugging Face, and Google Drive, at speeds the company says are 25 times faster than uploading from a local machine, and they persist between sessions. Separately, a playground serves hosted models such as Seedance, Wan, FLUX, and Nano Banana on per-generation pricing, which is a cheaper way to test a look before wiring up a workflow.
Templates and Team Sharing
More than 200 ready-to-run ComfyUI workflow templates cover image generation, video, upscaling, inpainting, and 3D work. Each one opens on the right hardware with dependencies resolved. Sharing a link hands a teammate the workflow, models, nodes, and settings in one step, which is handy for client handoffs and for teaching.
Pros and cons
Pros
- Uploading a workflow JSON and getting a runnable machine in minutes removes the most tiresome part of ComfyUI.
- The editor is unmodified ComfyUI, so graphs and habits carry over from a local install in both directions.
- Saved environments let you hold conflicting node versions at once without them interfering.
- The GPU range reaches 141 GB, which covers video and large open-weight models that consumer cards can't load.
- Billing runs by the second, and the serverless API scales to zero, so idle time doesn't show up on the invoice.
Cons
- The interface assumes you already understand nodes. Nothing here turns ComfyUI into a prompt box for beginners.
- GPU time adds up on long sessions. Rendering video for hours on an 80 GB machine at $4.99 per hour costs more than a year of a consumer subscription.
- Storage on the free tier is 10 GB and inactive assets are cleared after 90 days, so large model collections need a paid plan.
- ComfyUI is a fast-moving project, and workflows that depend on very recent node releases can need a manual fix even with auto-setup.
Frequently asked questions
RunComfy runs ComfyUI in the cloud. You supply a workflow, it builds a machine with the models and custom nodes that workflow needs, and you run the graph in a browser. It also turns saved workflows into API endpoints and hosts a separate playground for individual models.
Related content
Explore related tools, skills, and articles for RunComfy.
RunComfy Alternatives
Google EmbeddingGemma 2
Google DeepMind · Writing · Image · CodingGoogle EmbeddingGemma 2 is a 740M-parameter open model that turns text, code, images, audio, and video frames into vectors in one 768-dimension space, running on phones and laptops.
Tractionwave
TractionWave · Image · MarketingTractionWave is an AI attention heatmap tool for static ads. Upload a PNG or JPG and it returns predicted attention, ranked hotspots, and fixes before you spend budget.
Imejis.io
Imejis.io · Image · MarketingImejis.io is an AI design studio for agents and marketing teams. Describe an image in plain words, and it writes the copy, builds the layout, reflows it to every size, then exports a finished file.