Add MCP (Model Context Protocol) route support to @nestia/core

[Copilot]

This PR adds comprehensive support for MCP (Model Context Protocol) routes to @nestia/core, enabling developers to easily expose TypeScript functions as tools that can be called by LLMs through the standardized MCP protocol.

Features Added

Core Implementation

• @McpRoute decorator - Marks methods as MCP tools with configurable name, description, and JSON schema
• McpController base class - Automatically provides MCP protocol endpoints (POST /tools and POST /call)
• Type-safe tool definitions - Full TypeScript support for parameters and return types
• JSON Schema support - Manual schema configuration for tool parameter validation

Protocol Compliance

The implementation follows the MCP specification with:

• list_tools endpoint returning available tools with their schemas
• call_tool endpoint for executing specific tools with parameters
• Proper MCP response format with content arrays
• Standard error codes (-32601 for tool not found, -32602 for invalid parameters)

Usage Example

import core from "@nestia/core";

export class WeatherController extends core.McpController("weather") {
  @core.McpRoute({
    name: "get_weather",
    description: "Get current weather for a location",
    inputSchema: {
      type: "object",
      properties: {
        location: { type: "string", description: "Location to get weather for" },
        unit: { type: "string", enum: ["celsius", "fahrenheit"] }
      },
      required: ["location"]
    }
  })
  public async getWeather(params: { location: string; unit?: string }) {
    return {
      location: params.location,
      temperature: 22,
      unit: params.unit || "celsius",
      conditions: "sunny"
    };
  }
}

This automatically exposes:

• POST /weather/tools - Lists available tools
• POST /weather/call - Calls specific tools

Error Handling

• Comprehensive error handling with proper MCP error responses
• Tool-level exception catching with structured error messages
• Parameter validation and type checking

Testing

Added comprehensive test suite with 9 test cases covering:

• Tool discovery from multiple controllers
• Parameter validation and tool execution
• Error handling for invalid tools and parameters
• Real-world utility functions (UUID generation, Base64 encoding/decoding, JSON validation)

All tests pass, demonstrating full protocol compliance and robust functionality.

Integration with LLMs

The MCP routes can be used with any LLM that supports the Model Context Protocol:

• OpenAI GPT models (via MCP adapters)
• Anthropic Claude (native MCP support)
• Local LLMs with MCP implementations

Fixes #1350.

---

💬 Share your feedback on Copilot coding agent for the chance to win a $200 gift card! Click here to start the survey.