MCP Server
The CDS MCP Server provides direct CDS context to LLMs, allowing AI models to understand and work with CDS APIs and components.
From modelcontextprotocol.io:
MCP is an open protocol that standardizes how applications provide context to large language models (LLMs). MCP provides a standardized way to connect AI models to different data sources and tools.
First time setup
Run the setup command to install the CDS MCP Server for Cursor or Claude Code, then enable the MCP server in your agent.
Make sure to enable the MCP server after installation.
See the Cursor docs here and the Claude Code docs here for more information.
Using the MCP server
Once the CDS MCP Server is enabled in your agent, the agent will automatically be able to use the CDS MCP tools to interact with the CDS APIs and components.
Try asking a question like "How can I add an icon to a CDS Button?" or "How can I customize the CDS Stepper component behavior?"
Advanced installation
To install the MCP server separately as a standalone package and add it as a dependency to your project:
Or with Yarn:
Avoid installing mismatched versions of the CDS libraries and the CDS MCP Server. The MCP server is automatically included as a dependency of the CDS libraries.
Documentation
Documentation pages include "View as markdown" and "Copy page" buttons in a markdown like format. This is useful when pasting documentation into ChatGPT, Claude, or Perplexity for help with implementation.
Documentation indexes
CDS also provides a comprehensive index of documentation in markdown format per platform:
Web:
Mobile:
Each routes file contains links to all component documentation, React hooks, getting started guides, and brief descriptions for each item.
Adding indexes to Cursor
You can add these indexes to Cursor for searchable access to the entire CDS documentation library:
- Open Cursor Settings
- Navigate to "Indexing & Docs"
- Add the docs URL for your platform (web or mobile) using the routes links above
- AI will automatically fetch relevant docs when you reference components
Other AI IDEs and assistants can be given these indexes as needed or added to their rule files.