The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
/paper-sync, a Claude Code skill that keeps Paper's canvas and my React + Tailwind codebase in live bidirectional sync
Every prompt-to-UI tool I've tried hits the same wall: the last 10% where you describe every tiny visual tweak in chat. "Warm the green a touch." "Tighten the radius." "No, less." The AI regenerates the whole screen and you lose the thing you liked.
So I built the missing piece.
/paper-sync treats a Paper file as live data. I edit a color on the canvas with the picker, run the skill, and the right token in tokens.ts updates. The same run walks every mirror node in Paper, so the desktop and mobile artboards never drift. No describing. No regenerating. One pick, one sync.
How it works. Paper's MCP exposes every node and style as structured, queryable data. A snapshot file maps each design token to a canonical Paper node plus optional mirrors (other nodes that use the same value). The skill reads the canonicals, diffs them against the snapshot, applies a minimum code change (src/tokens.ts for token-bound fields, component files for one-off primitive styles), then reconciles the mirrors by pushing canonical values back into Paper.
Why it matters. Prompts can't adjust. They can only regenerate, and regenerating is lossy. Paper + MCP lets you do the adjustment visually and sync structurally in the same loop. That's the missing piece every AI design tool has left open.
30-second walkthrough: https://www.loom.com/share/f19bbead049d4b52827664f4811f07a1
Full writeup (argument, Paper vs Figma vs Claude Design comparison): https://igorgridel.com/blog/10-percent-nobody-ai-design-is-solving
Skill bundle (free for Patreon subscribers): https://www.patreon.com/posts/new-skill-paper-156143156
Built with Paper and Claude Code.
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Related posts
Using Claude code to model in blender helps to get the work done fast. Made this watch here in seconds using code generated by claude
What if the worst part of a creative project isn't the design, it's the brief that started it?
I built Dead Brief Resurrector on Melius — a workflow that takes a vague, broken client brief and turns it into a structured actionable direction, then immediately generates visual explorations from it. No manual prompting between steps. One input, full creative direction out.
The test brief I fed it:
"We need some nice pictures for our new luxury skincare line. It should look expensive and natural. Maybe some plants? Something that feels high-end but also very organic."
That's a real brief. Designers receive versions of this every week.
What the workflow did with it, 12 nodes, fully autonomous:
Ingested the vague brief as raw input
Ran a brief autopsy — identified every missing element: no audience, no tone, no visual direction, no KPIs
Resurrected it into a full structured brief: target audience (conscious luxury consumers), tone (serene, sophisticated, grounded), brand name (Gaia Essence), visual direction (Organic & Earthy Luxury)
Generated a 4-image moodboard across ingredients, environment, textures, and human element
Generated 3 distinct visual concept explorations: Brutalist Zen, Botanical Alchemy, and Terracotta Minimalism
My Melius experience:
The agent-first approach genuinely changes how you work. Instead of building nodes manually, you brief Mel the way you'd brief a creative director — in plain language — and it lays the canvas out for you. Every prompt is editable at the node level, so you can steer without rebuilding. The image quality surprised me, especially on the Botanical Alchemy concept. Where it still needs work: the agent occasionally needs explicit approval gates to chain text outputs into image nodes correctly — a small friction point in an otherwise smooth workflow.
Try the workflow yourself:
👉 https://app.melius.com/projects/091b9c6c-64ec-4baf-8ab9-39b80ed17daa/canvas/e8be693f-ec21-4d07-9851-5c4a7fa1eb6b
What brief would you feed it?
Great work
Hey Contra 👋 I've been heads-down building Promptly, and I'd love your eyes on it before our launch.
From my time delivering over 20 software projects as a tech lead, I realized the biggest bottleneck in building an app isn't the code itself. It is communication. Founders speak in ideas, and developers speak in strict logic.
Here is how Promptly works. You just describe your app idea in plain English. The AI actually pushes back, asks the hard questions, catches the vague bits, and automatically writes a developer-ready blueprint as you chat.
The goal is simple. When you hand this document to a developer, the next thing you get back is an accurate, fixed-price quote. No more wasting days trying to draft technical specifications and sitting through endless rounds of discovery calls just to get on the same page.
Instead of staring at a blank document, you get three tools that update in real-time:
The Pushback Chat: It refuses to just nod along. If you say "make it fast and secure," it helps you define exactly what that means. It hunts down edge cases before you ever pay someone to fix them.
The Visual Map: A live, interactive diagram showing exactly how your features connect. It catches logic flaws and missing steps before any code is written.
The Living Blueprint: An auto-generated, professional project spec that exports straight to Notion, Linear, or PDF.
Plus, it gives your idea a "Developer-Readiness Score" out of 100 so you know if it is actually ready to build. It even recommends the best tech stack based on your specific budget and complexity.
The core engine and all of these interactive features are fully functional right now, but I am still smoothing out the final rough edges. I am aiming for a public launch in about three weeks.
Where I need your help. I have three quick questions:
Where would you want to export your project blueprint, beyond Notion, Linear, or PDF?
What is the biggest headache you face right now when trying to explain your ideas to developers?
What feels missing? The blind spots are the most valuable to me right now.
Any feedback on new features, improvements, or things I should kill entirely would be hugely appreciated. Drop a comment below.
Thank you for taking a look 🙏
building solo, in public
Absolute killer problem to solve, that Visual Map alone is going to save founders thousands in wasted development hours.
Trending
Claude
Claude has entered the design space. How are you using Claude Design?
Contra University
Learn from expert creatives how to earn more using next-gen AI tools.
creativeaiflow
Creative AI workflows are evolving. What tools do you use, and what are their strengths and weaknesses?
portfolioreview
The best portfolios tell a story, not just show a grid. Share yours for feedback.
freelancerlife
Freelancer life is wins, pivots, and everything in between. What’s yours right now?