返回 MCP 目录
public公开dns本地运行

MCP远程服务

一个可配置的MCP服务器,它从远程配置动态加载功能,以桥接MCP客户端与远程API,从而执行操作、访问资源和使用提示模板。

article

README

Here's an enhanced README with more technical details based on the codebase:

mcp-remote-server

A configurable Model Context Protocol (MCP) server that dynamically loads its capabilities from a remote configuration. This leverages the ModelContextProtocol to create a local server that can communicate with your local MCP client.

Overview

mcp-remote-server acts as a bridge between MCP clients and remote APIs. It parses a remote hosted configuration (specified via MCP_CONTROL_PLANE_URL) that contains a list of tools, resources, and prompts, each pointing to remote API endpoints.

Key Features

  • Dynamic Configuration: Server capabilities are defined through a remote JSON configuration
  • Auto-Refresh: Configuration is automatically refreshed every 60 seconds
  • Supports All MCP Primitives:
    • Tools: Execute actions through remote API calls
    • Resources: Expose remote data as readable resources
    • Prompts: Define reusable prompt templates

Configuration Format

{
  "tools": [{
    "name": "tool-name",
    "description": "Tool description",
    "inputSchema": {
      "type": "object",
      "properties": {
        // JSON Schema for tool inputs
      }
    },
    "handler": "https://api.example.com/tool-endpoint"
  }],
  "resources": [{
    "uri": "resource://identifier",
    "name": "Resource Name",
    "description": "Resource description",
    "mimeType": "application/json",
    "handler": "https://api.example.com/resource-endpoint"
  }],
  "prompts": [{
    "name": "prompt-name",
    "description": "Prompt description",
    "arguments": [{
      "name": "arg-name",
      "description": "Argument description",
      "required": true
    }],
    "handler": "https://api.example.com/prompt-endpoint"
  }]
}

Handler API Requirements

Remote handlers must return appropriately formatted responses:

  • Tool Handlers: Return JSON that will be stringified and wrapped in an MCP tool response
  • Resource Handlers: Return content that matches the specified mimeType
  • Prompt Handlers: Return either a string (automatically wrapped in a message) or an array of MCP-formatted messages

Installation

bun install

Usage

  1. Set your control plane URL:
export MCP_CONTROL_PLANE_URL="https://your-config-endpoint"
  1. Run the server:
bun run index.ts

Limitations

  • Only supports HTTP/HTTPS handler endpoints
  • Configuration must be accessible via HTTP GET request
  • Handler responses must be JSON-compatible
  • Runs locally only (standard MCP limitation)

This project was created using bun init in bun v1.1.32. Bun is a fast all-in-one JavaScript runtime. I chose Bun because it's fast and I wanted to try something a little different.

This project also uses Biome for type checking and linting. I chose Biome for the same reason as Bun. I wanted something different and I wish JS had a gofmt-like tool. This is a good start.

help

运行方式说明

cloud

托管运行

托管运行通常表示这个 MCP Server 由服务方环境承载,用户一般按页面提供的连接方式或授权流程接入,不需要在本地长期启动一个 MCP 进程

  1. 打开服务方连接页
  2. 完成授权或复制端点
  3. 在 MCP 客户端中连接
terminal

本地运行 / 其它方式

本地运行通常需要用户在自己的电脑或服务器上安装依赖,把 server_config 复制到 MCP 客户端,并按 env_schema 补齐环境变量、密钥或其它配置

  1. 复制 server_config
  2. 安装所需依赖
  3. 补齐环境变量后重启客户端