Remote-MCP
A type-safe solution that enables remote access to Model Context Protocol (MCP) services, allowing clients to connect to centralized MCP implementations without waiting for official remote support.
README
Remote-MCP: Remote Model Context Protocol
A type-safe, bidirectional and simple solution for remote MCP communication, allowing remote access and centralized management of model contexts.
Architecture
%%{init: {"flowchart": {"htmlLabels": false}} }%%
graph TD
%% Modern, Bright Color Styling with white text
classDef client fill:#22c55e,stroke:#059669,stroke-width:2px,color:#ffffff
classDef gateway fill:#06b6d4,stroke:#0891b2,stroke-width:2px,color:#ffffff
classDef backend fill:#f97316,stroke:#ea580c,stroke-width:2px,color:#ffffff
classDef resource fill:#8b5cf6,stroke:#7c3aed,stroke-width:2px,color:#ffffff
classDef server fill:#06b6d4,stroke:#0891b2,stroke-width:2px,color:#ffffff
linkStyle default stroke:#64748b,stroke-width:1.5px,stroke-dasharray: 5 5
%% Current MCP Setup (Multiple Local Servers)
subgraph Current["Current Setup (Local)"]
direction LR
subgraph ClientGroup["Client"]
A[Client]:::client
end
subgraph Servers["Local MCP Servers"]
direction TB
B1["Local MCP Server (DB)"]:::server -->|"DB Access"| C1[DB]:::resource
B2["Local MCP Server (API 1)"]:::server -->|"API Access"| C2["Web API 1"]:::resource
B3["Local MCP Server (API 2)"]:::server -->|"API Access"| C3["Web API 2"]:::resource
end
A -->|"MCP Protocol"| B1
A -->|"MCP Protocol"| B2
A -->|"MCP Protocol"| B3
end
%% Vertical separator
Current --> Proposed
%% Proposed MCP Architecture (Decoupled)
subgraph Proposed["Proposed Architecture (Remote)"]
direction LR
D[Client/Host]:::client -->|"MCP Protocol"| E["Local MCP Server (@remote-mcp/client)"]:::server
E <-->|"tRPC(HTTP)"| F["Remote MCP Server (@remote-mcp/server)"]:::backend
%% Separated Resources
F -->|"DB Access"| G1[DB]:::resource
F -->|"API Access"| G2["Web API 1"]:::resource
F -->|"API Access"| G3["Web API 2"]:::resource
end
Why I Made This (Now)
Yes, I know that the official MCP roadmap includes remote MCP support in the first quarter of 2025. However, the need for remote access was immediate for me, and likely for many others. This library was created to bridge that gap, providing a way to connect to a remote MCP server from a local MCP client right now, without waiting for future official implementations.
Note: I don't want this to be a sophisticated or overcomplicated thing. This way just works right now.
Getting Started
Note: This project is currently under active development and is considered experimental. Expect breaking changes and potential issues.
Client Usage
Use Publicly Published Package
Just put the following code in your MCP client settings, in here I'm using Claude as an example:
{
"mcpServers": {
"remote-mcp": {
"command": "npx",
"args": ["-y", "@remote-mcp/client"],
"env": {
"REMOTE_MCP_URL": "http://localhost:9512",
"HTTP_HEADER_Authorization": "Bearer <token>"
}
}
}
}
Code Your Own Local MCP Server
Install requirements:
$ npm install @remote-mcp/client @trpc/client@next zod
then write your own code like the following:
import { RemoteMCPClient } from "@remote-mcp/client";
const client = new RemoteMCPClient({
remoteUrl: "http://localhost:9512",
onError: (method, error) => console.error(`Error in ${method}:`, error)
});
void client.start();
Server Usage (Remote MCP Implementation)
You can see some examples in the examples directory.
Code Your Own Remote MCP Server
After npm install @remote-mcp/server, you can your own remote MCP server like the following:
import { MCPRouter, LogLevel } from "@remote-mcp/server";
import { createHTTPServer } from '@trpc/server/adapters/standalone';
import { z } from "zod";
// Create router instance
const mcpRouter = new MCPRouter({
logLevel: LogLevel.DEBUG,
name: "example-server",
version: "1.0.0",
capabilities: {
logging: {},
},
});
// Add example tool
mcpRouter.addTool(
"calculator",
{
description:
"Perform basic calculations. Add, subtract, multiply, divide. Invoke this every time you need to perform a calculation.",
schema: z.object({
operation: z.enum(["add", "subtract", "multiply", "divide"]),
a: z.string(),
b: z.string(),
}),
},
async (args) => {
const a = Number(args.a);
const b = Number(args.b);
let result: number;
switch (args.operation) {
case "add":
result = Number(a) + b;
break;
case "subtract":
result = a - b;
break;
case "multiply":
result = a * b;
break;
case "divide":
if (b === 0) throw new Error("Division by zero");
result = a / b;
break;
}
return {
content: [{ type: "text", text: `${result}` }],
};
},
);
const appRouter = mcpRouter.createTRPCRouter();
void createHTTPServer({
router: appRouter,
createContext: () => ({}),
}).listen(Number(process.env.PORT || 9512));
Then you can see like the following in your MCP client:
<img src="https://github.com/user-attachments/assets/86cf500e-b937-47fc-9ac1-db106ab7a6a3" width="450">
Packages
This repository contains:
@remote-mcp/client: Client library acting as a local MCP server, connecting to a remote implementation.@remote-mcp/server: Server library for creating remotely accessible MCP services (used as the remote implementation).
Roadmap
Core Features
- [x] Basic Type-safe Client/Server Communication
- [x] Basic MCP Command Support
- [x] Basic MCP Tool Support
- [x] Basic MCP Prompt Support
- [ ] Crash-Safe Handling (WIP, top priority)
- [ ] Complete Event Subscription System
- [ ] Resource change notifications
- [ ] Tool/Prompt list change notifications
- [ ] HTTP Header Support
- [x] Custom Headers
- [ ] Authentication Middleware
- [ ] Basic error handling improvements
- [ ] Basic middleware support
Framework Support
- [ ] Nest.js Integration (
@remote-mcp/nestjs)
Advanced Features
- [ ] Bidirectional communication
- [ ] Server-to-client requests
- [ ] Resource sharing between server/client
- [ ] Basic monitoring & logging
Contribute
Contributions are welcome. See CONTRIBUTING.md for details.
Disclaimer
This library is a complementary extension, not part of the official MCP specification, built upon existing MCP concepts.
License
This project is licensed under the MIT License. See the LICENSE file for details.
References
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。