The official local Comfy MCP just shipped on August 18, 2026. It works with Claude, Codex, Cursor, and other MCP-compatible AI assistants. The MCP is one of the most interesting AI workflow updates I have come across lately. It can remove much of the technical setup that makes ComfyUI feel intimidating, so the focus stays on the creative decision: what do we want this image or video to feel like? ComfyUI is basically the engine room behind image and video generation. MCP gives an assistant a way to work with that engine room while you stay focused on the creative side: the shot, the reference image, the motion, and what is actually worth keeping. What MCP can help with: 🧠 Build a ComfyUI workflow from a creative brief ⚙️ Run workflows and help troubleshoot missing models or nodes 🧩 Adjust prompts, reference images, aspect ratio, duration, seeds, and settings 🎬 Connect different image and video models in one repeatable process 🔄 Help refine a workflow instead of rebuilding it from scratch each time This makes ComfyUI less about manually connecting boxes and more about having a flexible creative engine behind your work. You could use LTX for fast motion tests, then use a supported Seedance or MiniMax workflow when a shot needs a stronger final pass. The goal is not loyalty to one platform. It is having your references, prompts, models, and best settings working together. The bigger opportunity is building an advanced workflow once, then not having to actively engage with MCP or the individual ComfyUI elements for every generation. The workflow can sit behind a simple creative interface: a prompt field, reference-image drop zone, a few sliders, and a Generate button. You only return to the MCP or node graph when you want to change what is happening under the hood.

Hey everyone, I've been running an ICM folder system for my own reporting workflow and it's genuinely changed how I stay on top of everything. Now I want to roll it out to my immediate team first, then the wider company, but I want to do it in small increments rather than dropping it on everyone at once. Has anyone gone through this? Taking this system and actually getting a team or business to adopt it without it dying in week two? I'm thinking start with just my team, let them get used to the structure, then bring in other departments once there's some proof of life. But I'd love to hear from someone who's already been through the rollout side of things. What worked, what didn't, what you'd skip if you were starting again. Also, is there a video on here I should have watched already? If there's something that covers this and I've missed it, please just point me at it. Happy to do the homework first. Thanks

context I am currently working on an educational website, a platform featuring videos ,animations, documents etc Here is my current setup : point my ai agent (claude right now ) to the docs/ folder to understand the project and build a mental map to start working on the next goal (see screenshot N1) -I am a single "developer" working on this project with 0$ budget -Static front-end +dynamic backend (php) From what I understand ICM consists of simple markdown files and folders structured to function as an orchestration layer I know I did not provide enough context but just based on my current situation what advice do you have for me ? does ICM Fits My Project ?

I’m an AI full-stack engineer with 8 years of experience, and I’m currently looking to connect with agencies or individuals who can introduce me to potential clients. Many people have great ideas or business opportunities but hesitate to move forward because they lack the software development expertise needed to bring them to life. That’s where I can help. If you know someone who needs support with AI, software development, or turning an idea into a working product, I’d be happy to connect and explore how we might collaborate. Please feel free to reach out. I look forward to meeting new people, exchanging ideas, and exploring opportunities to work together!

Are you aware of this with GPT-Live and Projects? I’ve been Using GPT-Live daily to shape & complete so many tasks. The live conversation flow is getting closer to what people keep talking about here: voice as the interface of the future. But there’s one workflow gap that feels surprisingly clunky right now. You can start voice chats around Projects, and you can add files, folders, and other context into the voice conversation. But in practice, it still feels like the voice session becomes its own new conversation instead of just running inside the existing Project thread you were already working in. That means it may not automatically carry the same project-specific files, custom instructions, and prior context the way a normal Project chat does. So instead of simply switching an existing Project conversation into voice, you end up rebuilding context or starting a parallel conversation. That’s not a dealbreaker, but it does feel like friction in exactly the place where voice should make things smoother. Given how much we’ve talked in here about voice becoming the future interface for working with agents, I’m curious if anyone else has noticed this. My hope is that GPT-Live eventually becomes a true mode inside an existing Project chat, not a separate voice-first lane next to it. I still think the direction is right. I just hope this part changes, because Project-native voice would be a serious workflow unlock.