Agents
There are two ways to give AI tools access to the Meteor documentation. The MCP server provides a live, tool-driven connection that assistants such as Claude Code, Cursor, and VS Code can browse on their own. The llms.txt and raw markdown files expose the same content as static URLs you can paste straight into a prompt.
MCP server
The Model Context Protocol (MCP) is a standardized way for AI assistants to access external data and tools. Meteor ships an MCP server that lets assistants read the documentation directly and query structured data about components, examples, icons, and design tokens, so answers and generated code stay grounded in the current docs instead of model memory.
The server uses HTTP transport and is available at:
https://meteor.shopware.com/mcp
Tools
| Tool | Description |
|---|---|
list-pages | List all documentation pages with their paths and descriptions. |
get-page | Retrieve the full content of a documentation page. Component pages include the API reference and example code inlined as markdown. |
Resources
Browsable catalogs the assistant can read in a single call:
| Resource | Description |
|---|---|
meteor://components | Every component with its name, slug, and description. |
meteor://examples | Every documented example and its owning component. |
meteor://icons | The full icon kit (name, mode, tags). |
meteor://tokens | Design tokens grouped by category, each with its description. |
Configuration
Add the server to your AI assistant using the tab for your tool.
claude mcp add --transport http meteor https://meteor.shopware.com/mcp
{
"mcpServers": {
"meteor": {
"type": "http",
"url": "https://meteor.shopware.com/mcp"
}
}
}
{
"servers": {
"meteor": {
"type": "http",
"url": "https://meteor.shopware.com/mcp"
}
}
}
{
"mcpServers": {
"meteor": {
"command": "npx",
"args": ["mcp-remote", "https://meteor.shopware.com/mcp"]
}
}
}
Usage
Once the server is connected, you can ask your assistant questions like:
- "List the Meteor documentation pages"
- "Show me the mt-button documentation from Meteor, including its props"
- "Which Meteor component should I use for a confirmation dialog?"
- "What does the Meteor migration guide say about replacing sw- components?"
The assistant calls list-pages to discover what exists and get-page to read a page (component pages include the API reference and example code inlined as markdown). The meteor://components, meteor://examples, meteor://icons, and meteor://tokens resources let it browse the full catalogs directly.
llms.txt and raw markdown
llms.txt is a convention for exposing documentation in a format optimized for large language models. Instead of parsing rendered HTML, an AI tool can read a single structured text file that links to the full documentation as clean markdown.
Meteor publishes two files:
| File | Description |
|---|---|
/llms.txt | A concise index of the documentation with links to each page. |
/llms-full.txt | The complete documentation inlined into a single file. |
Start with /llms.txt for most cases. Use /llms-full.txt when the tool supports large context windows and you want the entire documentation set in one request.
Raw markdown
Every documentation page is also available as raw markdown by prefixing the path with /raw and adding a .md suffix. For example:
https://meteor.shopware.com/raw/documentation/getting-started/migration.md
Dynamic content such as component prop tables is rendered to plain markdown in these files, so the output stays useful to tools that cannot execute the site.
Using it with AI tools
Most assistants let you reference a URL directly in a prompt. Paste a link to /llms.txt or a specific /raw/*.md page and ask your question, for example:
- "Using https://meteor.shopware.com/llms.txt, which component should I use for a confirmation dialog?"
- "Summarize https://meteor.shopware.com/raw/components/button.md"