satori-mcp-server: Instructions file for Claude Code

This is a Satori MCP (Model Context Protocol) server that generates images from React components. Built with the xmcp TypeScript framework for creating AI tools with Model Context Protocol support.

Development Commands

# Install dependencies
pnpm install

# Run development server with hot reload
pnpm dev

# Build for production
pnpm build

# Start production HTTP server
pnpm start

Architecture

Framework Structure

• xmcp: TypeScript-first framework for building MCP applications
• Auto-discovery: Tools automatically discovered from src/tools/ directory
• Transport: HTTP transport configured via xmcp.config.ts
• Build Output: Compiled to dist/ directory

Tool Structure

Each tool in src/tools/ must export:

1. schema: Zod schema defining input parameters
2. metadata: Tool configuration with name, description, and behavior hints
3. default function: Async or sync implementation returning tool response

Tool Response Format

Tools should return either:

• Plain string/object (automatically wrapped)
• MCP-compliant response: { content: [{ type: "text", text: "..." }] }

Adding New Tools

Create a new .ts file in src/tools/:

import { z } from "zod";
import { type ToolMetadata, type InferSchema } from "xmcp";

export const schema = {
  paramName: z.string().describe("Parameter description"),
};

export const metadata: ToolMetadata = {
  name: "tool_name",
  description: "Tool description",
  annotations: {
    title: "Tool Title",
    readOnlyHint: true,      // Tool doesn't modify state
    destructiveHint: false,   // Tool is safe to run
    idempotentHint: true,     // Multiple calls yield same result
  },
};

export default async function toolName(params: InferSchema<typeof schema>) {
  // Implementation
  return {
    content: [{ type: "text", text: "result" }],
  };
}

Deployment

Vercel Deployment (Zero Configuration)

# Deploy to Vercel
vc deploy

# The framework supports Vercel deployment out of the box