Comfy Agent is an AI teammate built directly inside ComfyUI, the node-based canvas for image and video generation, for creators who want to describe a workflow in plain language instead of wiring every node by hand. It plans, builds, and runs ComfyUI workflows from a single prompt.
Core Features
- Plans, builds, runs, and iterates ComfyUI workflows from a text prompt inside the app.
- Up to five parallel agent chats with full history, plus reusable public or private skills.
- Real-time co-editing: the agent works on your canvas while you keep editing, without blocking each other.
- Visually understands assets and nodes, explains slow workflows, and fixes errors.
- Ships on Comfy Cloud today, with local ComfyUI agent support arriving next.
Use Cases
- Product photographers generating 20 on-brand ad variants across aspect ratios.
- VFX artists isolating, masking, and refining shots into compositing passes.
- Animation studios shaping character looks and comparing model outputs.
- Teams onboarding to ComfyUI without learning every node.
Pricing
Free to try using the Comfy Credits you already hold in Comfy Cloud, with no separate subscription during the beta. Paid Comfy plans unlock the Comfy API and higher throughput, while the coming local agent mode runs on your own GPU at no per-use fee.
Our Take
Best for visual creators who live in ComfyUI and want its control without the wiring labor. The catch is that it is beta and cloud-first for now, and fully local custom-node authoring is still on the roadmap. Explore related builders in our AI Design category.
