Every site gets a public MCP server that AI platforms can query — no API keys, no configuration needed.
What is the MCP Client?
@webhouse/cms-mcp-client is a public, read-only MCP server bundled with every @webhouse/cms site. Any AI platform — Claude, ChatGPT, Cursor, Copilot — can discover and query your published content without API keys.
When someone asks an AI "What does webhouse.app do?", the AI can connect to your MCP endpoint and read your actual content to formulate an accurate answer.
6 Tools
| Tool | Description |
|---|---|
get_site_summary | Site overview: name, description, language, collections, document count |
list_collection | List published documents in a collection (limit, offset, sort) |
search_content | Full-text search across all published content |
get_page | Get full content of a page as markdown + metadata |
get_schema | Field schema for a collection (field names, types) |
export_all | Export all published content as structured JSON |
Setup
The MCP client is automatically available when your site runs. No configuration needed.
For Claude Desktop or Cursor, add to your MCP config:
{
"mcpServers": {
"my-site": {
"command": "npx",
"args": ["@webhouse/cms-cli", "mcp"]
}
}
}Or connect via SSE transport to a running site:
https://your-site.com/api/mcpTool details
get_site_summary
Returns site name, description, default locale, all collections with document counts.
No parameters required.
→ { name, description, locale, collections: [{ name, count }], lastBuild }list_collection
{
collection: "posts", // required
limit: 20, // optional, max 100
offset: 0, // optional
sort: "date_desc" // optional: date_desc, date_asc, title_asc
}Returns published documents with title, slug, excerpt, date. Content body excluded for performance.
search_content
{
query: "typescript generics", // required
collection: "posts", // optional: scope to one collection
limit: 10 // optional, max 50
}Full-text search across all fields. Returns matching documents with relevance score.
get_page
{
slug: "getting-started", // required
collection: "docs" // optional: scope the lookup
}Returns full markdown content + all metadata fields.
get_schema
{
collection: "posts" // required
}Returns field definitions: name, type, required, options. Useful for AI agents that need to understand content structure.
export_all
{
include_body: true // optional, default true. false = metadata only
}Exports everything. Use with caution on large sites.
Rate limiting
The public MCP client includes built-in rate limiting to prevent abuse. No configuration needed.
Why this matters
Traditional CMS platforms require API keys and documentation for AI access. @webhouse/cms makes your content discoverable by default. AI platforms can read your content, understand your schema, and cite your pages accurately — all without you setting anything up.