The whole library,
inside your AI editor.
What is MCP?
MCP, the Model Context Protocol, is an open standard that lets AI assistants use tools from outside services. An assistant that supports it can call those tools in the middle of a conversation, the way it reads a file or runs a command.
The Cortexcn MCP server gives your assistant read-only tools over the Cortexcn registry. It can search for components and blocks, read an item's docs and source, and get the command that installs it.
How it helps
Finds the right piece
Describe what you need, and it searches components and blocks by name or by what they do.
Reads before it installs
It sees each item's license, revision, files, and npm packages, so you can review the change first.
Installs the source
It gets the exact shadcn command, your assistant runs it, and the files land in your repo.
Keeps your style
It reads Cortexcn's DESIGN.md, so new code follows the same tokens as the rest of your UI.
- HostedNothing to install or run
- Read-onlyIt never writes files or runs commands
- No sign-inOne URL for every client
- Fair use120 requests a minute per IP
Install the MCP server
Pick your assistant. Every client connects to the same URL, and there is nothing to sign in to.
Run this in your terminal:
claude mcp add --transport http cortexcn https://ui.cortexcn.dev/mcpAdd --scope user to use it in every project. Run /mcp inside Claude Code to check that it is connected.
Using Zed, ChatGPT, or another client? The MCP reference covers them all.
Then ask for what you need
Your project needs shadcn set up first. If there is no components.json in its root, run npx shadcn@latest init.
"Find a hero block with a product screenshot and add it to the landing page."
"Add the feature-bento-grid-01 block under the hero."
"What props does the Cortexcn dialog take?"
"Show me the source of hero-cover-01 before we install it."
Every tool, limit, and client
The reference covers the eight tools, the rate limits, and setup for Zed, ChatGPT, and more.