Aicanvas

What aicanvas MCP server does

The aicanvas MCP server gives an MCP-compatible AI editor access to the AI Canvas catalog. The catalog contains animated React components and composed page blocks intended to be added to a project as source code. Examples in the catalog include task boards, galleries, card decks, signature inputs, and other interactive UI pieces.

The server exposes the catalog to an agent rather than acting as a runtime UI dependency. An editor can ask it to find relevant entries, inspect their published details, and retrieve source that can be used in the project. This suits developers who want an AI coding tool to select existing UI implementations instead of recreating them.

How it works

The aicanvas MCP server is launched as a local stdio process with npx:

npx -y @aicanvas/mcp

An MCP client starts the process and communicates with it through the client’s MCP configuration. At request time, the server fetches the live AI Canvas registry, so newly published components can become available without requiring a local catalog update.

The server is read-only. Its role is to return published component metadata and source; the connected agent or developer can then apply that source to a project. This is separate from the authenticated shadcn registry installation flow described by AI Canvas. The MCP connection itself does not require a documented environment variable in the provided setup.

Setup and configuration

For Claude Code, the README gives this command:

claude mcp add aicanvas -- npx -y @aicanvas/mcp

Alternatively, add the server to an MCP configuration file:

{
  "mcpServers": {
    "aicanvas": {
      "command": "npx",
      "args": ["-y", "@aicanvas/mcp"]
    }
  }
}

The README names Claude Code, Codex, Cursor, and other MCP-compatible editors as supported clients. No API key or server-specific environment variable is listed for this MCP setup. AI Canvas does require an account token for direct authenticated shadcn CLI installs, but that is a separate workflow and should not be confused with the MCP server configuration.

Tools and capabilities

The aicanvas MCP server supports these catalog-oriented actions:

• Search for components and blocks in the live registry.
• Inspect published component metadata.
• Retrieve published component source.
• Help an AI editor identify components to install into a project.

The returned material represents AI Canvas’s free component library and its composed blocks. The source is React and TypeScript, and the library is written for Tailwind CSS v4, React 19, and Framer Motion. Some three-dimensional pieces use Three.js. These project requirements come from the catalog and are relevant when deciding whether an identified component fits an existing application.

Limitations and notes

The aicanvas MCP server is read-only, so it does not modify a repository by itself. An agent may use the returned source in an installation or editing workflow, but the server’s documented responsibility is retrieving registry content.

The catalog components target Tailwind CSS v4 rather than Tailwind CSS v3. React 19 and Framer Motion are also listed requirements. The Next.js App Router is the default target, although it is not required. Free components are MIT licensed according to the README; premium components, design systems, and templates are described separately as proprietary.

Direct shadcn installation requires an AI Canvas account token and registry configuration, while reading component source from the site does not. The MCP documentation provided here does not specify separate authentication, write operations, version pinning, or offline registry access.