Intelligent
Knowledge Core
A hosted MCP server combining semantic RAG search with precise code grepping over the SveltyCMS codebase.

Real-time Retrieval
High-speed vector search serving all apps. One core serves the entire monorepo with millisecond responses.
Ecosystem Driven
Built-in Model Context Protocol support. Unlocks professional-grade context for any MCP-compatible AI agent.
Static Wisdom
Pre-indexed memory shipped to the cloud. Optimized for remote retrieval with zero GPU overhead on the server.
Connect Your Agents
The Knowledge Core is accessible via Streamable HTTP (MCP spec 2025-03-26). Follow the guides below to integrate it into your development workflow.
VS Code / Cursor / Zed / Claude
Best for Cursor, VS Code (Cline/Roo Code), and Zed which support native Streamable HTTP connections.
Add this to your MCP configuration file:
{
"mcpServers": {
"sveltycms": {
"serverUrl": "https://mcp.sveltycms.com/mcp"
}
}
}
Custom Bots & SDKs
Perfect for building automated agents or custom CI/CD memory tools.
import { StreamableHTTPClientTransport } from "@modelcontextprotocol/sdk/client/streamableHttp.js";
const transport = new StreamableHTTPClientTransport(
new URL("https://mcp.sveltycms.com/mcp")
);
await transport.start();
// Connect and interact with tools...
Available Tools
All tools are read-only. Your agent can explore the SveltyCMS codebase but cannot modify anything.
search_knowledge_baseSemantic search across SveltyCMS docs and code via neural vector similarity.
read_source_codeRead the full content of any source file from the SveltyCMS project.
list_directoryList files and folders in a given directory.
get_file_treeGet a recursive tree of the project structure up to a specified depth.
search_codeFast text/regex search across all project files. Finds component imports, function references, and symbol usage.
list_collectionsList all content collections with field definitions and metadata.
query_contentQuery content entries with filters, sorting, and pagination.
create_content_itemCreate a new content entry in any collection.
update_content_itemUpdate an existing content entry with partial data.
delete_content_itemDelete a content entry by its ID. Use with caution.
Available Resources
MCP Resources represent static data or documentation that agents can read directly.
sveltycms://docsProvides the complete SveltyCMS developer handbook, outlining component architecture, state management, and runes conventions.
Available Prompts
Prompts are one-shot templates that guide the AI through specific tasks. Select them from your AI assistant's prompt picker.
analyze-componentDeep-dive analysis of a Svelte 5 component: props, runes, reactivity, accessibility, and CMS integration.
find-unused-codeScan a file or directory for dead code — exports never imported, functions never called.
svelte-4-to-5-migrationAnalyze a Svelte 4 component and generate a line-by-line migration plan to Svelte 5 runes.
code-reviewComprehensive review: TypeScript types, Svelte 5 patterns, a11y, performance, security, and CMS conventions.
Safe & Read-Only
This MCP server is completely read-only — no write tools exist. Your agent can explore the SveltyCMS codebase, search documentation, and read source files, but it cannot modify, delete, or create anything.
Sensitive paths are blocked: .git, node_modules, .env, config, logs
Example Queries
Try these questions in any MCP-compatible AI assistant:
search_knowledge_base "How does the Local SDK caching work?"
search_knowledge_base( query: "How does the Local SDK caching work?", category: "docs", limit: 5 )
read_source_code "Show me the Local SDK entry point."
read_source_code( filePath: "src/services/sdk/index.ts" )
search_code "Where is input validation implemented in the services?"
search_code( pattern: "validate", includePattern: "src/services/**", maxResults: 20 )