
Remote-MCP Server
Enables type-safe, bidirectional communication with Model Context Protocol services, allowing centralized management of model contexts over HTTP.
ssut
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
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Playwright MCP Server
Provides a server utilizing Model Context Protocol to enable human-like browser automation with Playwright, allowing control over browser actions such as navigation, element interaction, and scrolling.
MCP Package Docs Server
Facilitates LLMs to efficiently access and fetch structured documentation for packages in Go, Python, and NPM, enhancing software development with multi-language support and performance optimization.
Claude Code MCP
An implementation of Claude Code as a Model Context Protocol server that enables using Claude's software engineering capabilities (code generation, editing, reviewing, and file operations) through the standardized MCP interface.
@kazuph/mcp-taskmanager
Model Context Protocol server for Task Management. This allows Claude Desktop (or any MCP client) to manage and execute tasks in a queue-based system.
Apple MCP Server
Enables interaction with Apple apps like Messages, Notes, and Contacts through the MCP protocol to send messages, search, and open app content using natural language.
contentful-mcp
Update, create, delete content, content-models and assets in your Contentful Space