正在加载视频...
视频加载失败
Comfy Agent is now available to everyone in Comfy Cloud. It plans, builds, and fixes ComfyUI workflows right on your canvas. We strongly believe creators shouldn’t choose between ambition and complexity. Comfy Agent is our answer. While it handles the technical part of Comfy, you can finally focus on... show more
36 条评论

Try Agent today:

Learn more with our blog:

Congratulations on the release! I've been having a lot of fun working with it. thanks for the early access ❤️🔥 excited to keep going with it

killer release and big missing piece here!

This is probably the biggest day if we look back in 5 years!

THIS TEAM CAN'T STOP SHIPPING!

Such a great addition to comfy!

The new best way to learn comfyui 🤩

So excited to see this out in the world now!!

you guys can't stop shipping! 🤯

Does it explain what it changed when it fixes a workflow? I'd want to know why a node moved before I trust it on a client job.

Curious to see what creators do with this once they start pushing it beyond the obvious use cases. The workflow possibilities are pretty wide.

An agent that fixes workflows right on the canvas, node-graph nightmares finally have an answer.

I really like the human + agent setup here. The AI handles the tedious parts, but the creator still has full control.

让我逆向一下他的prompt

This is the kind of agent experience that actually makes sense. It’s working inside the tool rather than sitting beside it.

I really like that the underlying workflow stays visible and editable.

This is a good example of AI assisting with the process, not just producing the final output.

This could make some of the more intimidating parts of ComfyUI much easier to approach.

seeing the same assets as the creator makes the agent much more useful for visual work

This feels especially powerful for repetitive creative workflows.

I’m particularly curious about what people build with this once they start pushing it beyond basic workflows.

The idea of an agent that understands models, nodes, extensions and your actual canvas is pretty powerful.

This is a much more practical way to think about AI agents for creative work.

I like that creators still get the final say over what actually happens.

The iteration part is probably where this becomes really useful.

The interesting question now is how far creators can push these agents once they become part of the normal workflow.

This makes ComfyUI feel much more approachable without taking away the control.

The combination of creative control and agent automation feels right for a tool like ComfyUI.

Does the asset awareness survive a checkpoint swap mid session?

I like that ComfyUI’s underlying control isn’t going away. The agent is adding a faster layer on top of it.

Being able to test multiple models without rebuilding everything manually could save a ridiculous amount of time.

This could turn a lot of repetitive ComfyUI work into a simple conversation.

The fact that it can understand assets visually makes this feel much more native to a creative workflow.

A lot of creative work is iteration. Anything that makes iteration cheaper is worth paying attention to.

Having an agent build and fix workflows directly on the canvas could remove a huge amount of technical friction for creators.
