Inspire me

I finally have the time to test the new Framer AI Agent feature, and I'm really impressed! Most of you don’t know what a lifesaver this update is for my current real workflow! What used to take me 1–2 hours of manually uploading CMS content now takes just 5 minutes, while I’m off doing something else. Previously, I used Claude together with the MCP plugin from the marketplace. It got the job done, but it would sometimes make mistakes, which meant I had to hire someone to double-check everything in the CMS, or I had to spend some time checking it myself, which I don't want to. There were also limitations when it came to uploading images. Now that Framer has its own MCP integration, the agent experience feels much more polished. The agent operates within the context of the project, which means it's much less likely to go off-script. Here's my current workflow: I have Claude connected to my Notion database, where I store all the websites I curate for agencygallery.com . I also told Claude to read my Google Drive folder. 1. I give Claude a screenshot of my Notion list with the websites I want to add to Framer. 2. It finds the matching folder in Google Drive using the same name as the Notion entry. It uploads the screenshots from that folder to Framer. 3. It adds the website to Framer, following the same pattern as the existing CMS items, across 2 different CMS collections. 4. Once everything is added, the status on Notion is updated automatically. 5. Then I ask it to generate image alt text for all the uploaded images. It nailed everything in a single shot, in sonnet 4.6. Such a huge time saver!!!!!

Last week Framer 3.0 was released, the biggest update the platform has shipped in years. AI Agents now live right on the canvas, so you can ask them to design pages, wire up the CMS, build components, add effects and write code without leaving your project. So now I can finally talk about what we prototyped during the hackathon: agents interacting with Framer shaders. In the attached video I asked the agent to build a portfolio for my shader work. Once it laid out the structure and a first iteration, it read the actual shader, understood the current setup, and came back with multiple-choice questions: which palette to use, how the gradient should move, how strongly the letters and background should interact. The agent never touched GLSL. It didn't generate new shader code for me to read through and audit for performance. It designed using the controls I'd already exposed while building the shaders, with the same optimized code running underneath. I get fast iteration on the look with zero risk to how the shader performs. And /ask-me flips the usual flow: the agent interviews me about what I want before it commits to anything. This opens up a different way of designing in Framer. Setup work that used to eat hours now takes minutes, which leaves far more room for the creative, experimental side of the work. Branching makes that even safer: spin up a separate branch, try the wild ideas with no fear of breaking the live project, and once something is polished you commit it back to main. Curious what you'd build with it.

I’ve been putting Framer ’s AI agents to work on the product page I set up in Swell and synced into Framer through The Good Commerce . I gave it a single prompt and five minutes later had a three-column product layout—image centre, description left, title/variants/price fully built and bound to real data. Normally this is a few hours to days of work, from designing the layout, each component, building each property, then connecting all of it to the data by hand but the agent was able to do it in one pass. I'll still fine tune any details manually, like the margins, the typefaces, the personal touches. That's the part I want to be on. It took about five minutes to do something that usually takes me an afternoon. I'm seriously impressed by Framer's AI agents. They remind me of the first time I started using AI. I'm excited to see what else I can do with them, and what Framer does next. #framerpartner #framer

This morning I had 70 whitepaper CMS pages to build for a client Not hard. Just boring and repetitive So I gave the new Framer agent a try, they launched it last week What used to take an hour of grunt work? done in 30 seconds The lead magnet template was already built, with every field mapped Populating it was the part I dreaded Each page needed real content. My old process meant: → Open Claude Code, pull the content for each one → Format it to fit the template → Paste every field into Framer by hand → Repeat until the morning's gone A few minutes a page, it adds up fast This time I opened the Framer agent instead, right inside the CMS collection No links pasted. No setup. Just asked it to fill the empty items 30 seconds later: → It fetched the URLs on its own → Read each whitepaper, filled every field → Pages ready to deploy And this is just the starting point Not AI slop that looks busy and adds nothing. Real work, with real business value These are the tasks AI can make vanish That's the pattern now. You bring the system. The agent brings the speed For a B2B team shipping content weekly, that's hours back. Every week Thinking about switching to Framer? We can help Comment below and I'll reach out personally to tell you honestly if it's the right move for you. See everything in Framer 3.0 at https://www.framer.com/

Design world-class sites with Framer Agents: prompt, refine, and ship production websites without leaving your canvas.

My favorite skill in the Framer agent is /ask-me. Instead of guessing, it asks you for more context first.

Wondering how we built these interactive components on Framer.com ? Hint: We used Framer Agents. Prompt 1: 3D Image Wheel 🎡 /code Build a Framer code component that displays uploaded images in a draggable rotating ring that speeds up on scroll. Prompt 2: Before / After Image Slider 🖼️ /code Build a component that compares two images side-by- side using a draggable vertical divider to reveal the before and after states.

Framer's new update is out and everyone is talking about the new built-in AI agents. I'm impressed too. But for me the real magic is not in flashy landing pages — it's in how much boring routine work these agents can simply do for you. Check this screencast: updating these CMS images manually would take me 10x longer! That's exactly the AI help I needed 😌

Design world-class sites with Framer Agents: prompt, refine, and ship production websites without leaving your canvas.

I love that I can use Claude code much easier now to edit things for me in Framer , but dang it, the user experience and the output of Framer agents is wayyy too good.... e.g. being able to @ anything like components or pages is so clutch!!

Use the Framer agent to create color variants of your design, even with shaders.

Watch me use Framer AI Agents to build a new pricing page based on minimal context from a section on my homepage. This is the future of web building.

Design world-class sites with Framer Agents: prompt, refine, and ship production websites without leaving your canvas.

Last week Framer shipped its biggest AI update. The site builder got agents. That's the headline. But I discovered something interesting. I've spent the weekend in design pages, and the AI agent is doing the things I've actually wanted from a design tool for years. I had scattered designs across my canvas. Loose layers, inconsistent text, and components that had drifted into five slightly different versions. I asked the agent to clean it up. Match everything to the design system. Done in one pass. I selected the component variants and asked it to simplify and reconcile them. It did, and made the right calls about what to keep. Building components, setting type styles, and keeping a system consistent while you move fast. The slow manual parts. That's where the AI showed up. We've had a lot of announcements about AI in design. Not many real use cases. This is the first one that felt like it was built for how I actually work. It's early. I know that. But I'm looking forward to designing more in Framer, not just publishing a site out of it. #FramerPartner

This line from Andy Orsow during the Framer 3.0 launch stayed with me: “Design with Agents. Refine on the Canvas.” I think it captures the right way to think about AI in a design workflow. Not as a replacement for the creative process, but as a new collaborator inside it. The interesting part of Framer Agents is not only that they can generate a new page from a prompt. It’s that they can work inside the same canvas where a real production site is designed, maintained, reviewed, and shipped. That matters because a lot of web design work is not just the first concept. It is the repetitive and less visible work around it: updating sections, creating a styleguide, fixing inconsistencies, testing a different direction for a hero, cleaning up layouts, checking responsive behavior, managing content, or making larger batch edits without losing control of the system. As Joseph Angelo Todaro said during the launch, this is the “tedious and repetitive stuff” where an agent can jump in. The designer still defines the direction, the hierarchy, the visual language, and the final decision. The agent helps reduce friction around the work that slows the process down. What I also like is that everything the agent creates remains editable and controllable on the canvas, as if you had designed it yourself. It can start from the structure, components, and decisions already present in the project, which makes the process feel collaborative rather than detached. Branching makes this even more useful for real teams. You can create a safe little sandbox, explore a new direction, test a larger set of changes, invite teammates into that branch, discuss specific decisions, and only merge when the work is ready. For me, this is the strongest part of Framer 3.0. Agents bring speed and scale. The canvas keeps craft and control in the designer’s hands. Branching makes experimentation safer for production work. That feels like a more mature way to bring AI into web design. #Framer #WebDesign #AIDesign

I used to spend 1 to 2 hours uploading CMS content in Framer . Now it takes 5 minutes using Framer Agents with Claude. Not because I got faster. Because I stopped doing it myself. Framer's AI Agent has Claude inbuilt now and it understands the context. Your time is the most valuable asset in your freelance design business. My workflow: 1. I create a short document with the titles and description of the project. Then get Claude to format it and label each item with cms fields before and convert it into a PDF / Markdown text file. 2. I drop my project images and my doc into the agent chat in Framer. I ask it to create a CMS using the doc and images as the reference. 3. It automatically matches the images with each project, the content and fields. 4. It will create the CMS collection in Framer, and when I want to add a new project I just upload new images with the project description and it will match the pattern of the CMS we already created. 5. Then I ask it to generate image alt text for all the uploaded images. Sonnet 4.6 get's it done in one go, pretty wild. Are you using Framer AI Agents in your workflow yet?

Design world-class sites with Framer Agents: prompt, refine, and ship production websites without leaving your canvas.