AIDesigner
UI So Good You Can't Tell It's AI
Start new thread
Tyler

AI Designer MCP - Give your agent tools to create beautiful, codebase-aware UI

by•
Claude Code and Codex are great at writing code, but frontend design is still where coding agents fall short. That’s why we built the AI Designer MCP — giving your agent the tools to create beautiful, codebase-aware UI directly inside the client you already use. No more generic purple gradients, overused lucide icons, and AI-looking layouts. Just clean, polished, relevant UI that actually fits your product.

Add a comment

Replies

Best
Tyler
Maker
📌
Hi everyone, I'm Tyler, the founder of AI Designer! If you’ve been using Claude Code or Codex to build web apps, mobile apps, or product UI, you’ve probably noticed the same thing I did: they’re solid at coding, but pretty mediocre at design. I originally built AI Designer as a separate design platform to help with my own UI work, but there were always two big problems: 1. It didn’t have direct context of my existing codebase and design system. 2. I had to constantly bounce between tools — design in one place, export it, then bring it back into my coding workflow That back-and-forth got annoying fast. What I really wanted was for Claude Code itself to just be good at UI design — to create designs that actually fit the project I was working on, without forcing me to leave my coding environment. So I built the AI Designer MCP. It gives coding agents like Claude Code access to design tools and skills so they can generate beautiful, relevant UI and incorporate it directly into your codebase. I’ve been using it myself to create new pages and revamp existing ones, and it’s been a much better experience than using plain Claude Code alone. Setup is simple — one terminal command installs the MCP and skill files so your agent can start using it right away. You can learn how to it works and how to try it out here: https://www.aidesigner.ai/docs/mcp It’s free to try, and since this is a fresh release, I’d genuinely love feedback from anyone who gives it a shot. Really curious to see what people build with it!
Priya K

i was ready to ignore another ai design tool until i saw the part about encrypted secrets and local context. one quick question "does it pull from the global design system file automatically or do i need to point it to specific files? @bowlcutwiz @AIDesigner

Tyler
Maker

@priya_kushwaha1 Hey Priya! Yeah the MCP comes equipped with a command/instructions for your agent to create a design system markdown file based on your existing codebase. Your agent will reference that design.md when creating future designs.

It's not a requirement though, it's capable of creating designs from scratch and from analyzing individual pages/components as well.

Mykola Kondratiuk

skeptical about fully autonomous UI generation - every agentic system I run hits edge cases where the output breaks existing design contracts. what's the human review point before it commits changes?

Tyler
Maker

@mykola_kondratiuk Understandable concern and that's the beauty of an MCP! We're just giving the agent the tools to create good UI while the human remains the orchestrator and reviewer. All designs are still run by the user, adjusted however necessary before any commits happen.

In terms of staying consistent with existing design systems, because the agent has context of the user's existing codebase, it's actually quite good at creating UI that matches existing design structure. The MCP also comes equipped with skills that instruct the agent on how to properly construct consistent design prompts and wire outputs back into user's project.

Mykola Kondratiuk

fair distinction. tool-call model gives cleaner control than pure generation. still depends on how tight the review loop is before commit.

Tyler
Maker

@mykola_kondratiuk 100%. Reviewing design output is actually quite easy with the MCP since each design artifact gets stored as both raw code and as a preview image so both the agent AND the user can preview it before deciding to move forward with any changes.

Mykola Kondratiuk

That artifact structure makes sense - raw code + preview gives a proper review trail, not just output to eyeball. Lighter review step changes the math on autonomy.

Athsara Fernando

If I iterate multiple times on the same project, does it maintain consistency or drift in design decisions?