mcp-composer MCP Client

Use This MCP client To

Consolidate multiple MCP servers into a single access point Dynamically activate or deactivate MCP servers and tools Create independent SSE endpoints for different AI agents Customize MCP server combinations per AI application Manage MCP server connections centrally for easier orchestration Enable flexible AI app control over multiple MCP servers Simplify multi-server MCP workflows with unified SSE streaming

README

MCP Composer

UI Demo

MCP Composer is a gateway service that centrally manages all your MCP servers. You can use it to consolidate all the MCP servers you need and open independent ports with different combinations of servers and tools for each service (like AI agents or tools) that requires access to MCP servers.

Key Features

Dynamic MCP Server Management: Dynamically manages connections to multiple MCP servers and their tools, enabling on-the-fly activation or deactivation of services.
Unified SSE Interface: Exposes a single Server-Sent Events (SSE) interface that provides access to all capabilities of the managed MCP servers.
Multiple Dynamic Endpoints: Supports dynamic creation and removal of multiple SSE endpoints to accommodate different AI agents or AI tools.
Independent Interface Configuration: Each SSE interface independently manages its own combination of MCP servers and tools, allowing for customized service provision.

System Architecture

MCP Composer Architecture Diagram

Key Terms

MCP Client: External tools or AI agents and AI workflows, such as Cursor, n8n, etc.
Gateway[A/B]: MCP server implementation bound to ServerKit and managed by the Composer, used to handle MCP Client connections.
Server Kit: Manages and controls information about Downstream MCP servers and which tools within servers should be enabled.
Downstream Controller: Responsible for controlling and managing connections to Downstream MCP servers.
Downstream MCP Server: An internal object connecting to a Downstream MCP server, implemented as an internal MCP client that links to the Downstream MCP server.
MCP Server: External MCP server services, such as: https://github.com/modelcontextprotocol/servers etc.
Composer: Provides APIs to control and orchestrate Gateways, Server Kits, and Downstream Controllers.

Requirements

Python >= 3.12
uv: A fast Python package installer and manager.

mcp-composer FAQ

How does MCP Composer manage multiple MCP servers?

It dynamically connects to and controls multiple MCP servers and tools through a single SSE interface, allowing flexible activation and configuration.

Can I create separate endpoints for different AI agents?

Yes, MCP Composer supports multiple dynamic SSE endpoints, each independently managing its own combination of MCP servers and tools.

What is the benefit of using a unified SSE interface?

A unified SSE interface simplifies streaming data and events from multiple MCP servers into one consolidated channel for easier consumption by AI apps.

How does MCP Composer handle server activation and deactivation?

It allows on-the-fly activation or deactivation of MCP servers and tools, enabling dynamic service management without downtime.

Is MCP Composer suitable for complex AI workflows?

Yes, it is designed to orchestrate multiple MCP servers and tools flexibly, making it ideal for complex AI agent and app workflows.

Can MCP Composer be used to customize MCP server combinations?

Absolutely, each SSE interface can be independently configured with different MCP servers and tools tailored to specific AI services.

Does MCP Composer support scalability for multiple AI applications?

Yes, by creating multiple dynamic SSE endpoints, it supports scalable deployment for various AI agents or tools simultaneously.

What protocols does MCP Composer use for communication?

It uses Server-Sent Events (SSE) to stream real-time data from MCP servers to clients efficiently.