McpManager
Aggregate and expose multiple Model Context Protocol (MCP) servers through a single, self-hosted endpoint, simplifying access for AI clients.
McpManager acts as a central hub between your AI tools (such as Claude Code, Cursor, or Windsurf) and your various Model Context Protocol (MCP) servers. Instead of individually configuring each server in every client, you register them once within McpManager. This platform then aggregates tools from all registered upstream MCP servers and exposes them through a single, unified MCP endpoint, allowing AI clients to connect once and access a comprehensive set of tools without per-server configuration overhead. It streamlines the management, customization, and access to diverse AI agent capabilities.
주요 기능
012 GitHub stars
02Namespace Organization: Group MCP servers into logical namespaces with independent rate limiting and API key scoping.
03Unified MCP Proxy: Aggregate tools from all registered upstream MCP servers into a single endpoint.
04Multi-Transport Support: Connect to remote servers via HTTP/SSE, Stdio for local CLI tools, or auto-convert OpenAPI specs.
05Tool Customization & Control: Override tool names, descriptions, parameters, and enable/disable individual tools per server or namespace.
06API Key Management: Generate scoped API keys with optional namespace restrictions for authenticating MCP client connections.
사용 사례
01Centralizing access to various AI agent tools and services through a single unified endpoint for clients.
02Managing and customizing the tool sets exposed by multiple MCP servers to different AI clients or namespaces.
03Converting existing REST APIs from OpenAPI specifications into callable Model Context Protocol tools for AI agents.