Install guide and config at curatedmcp.com

MCPX - MCP Gateway: Secure, Route, and Control Your Entire MCP Ecosystem

Managing multiple MCP servers across your AI workflows shouldn't require hand-rolling security, routing logic, and permission layers. MCPX is a lightweight, open-source gateway that sits between your AI agents and your MCP services—handling routing, access control, and unified API exposure without the overhead.

Think of it as a smart proxy for your MCP stack. Instead of scattered integrations and manual security enforcement, you get a single control plane that routes requests intelligently, enforces permissions, and lets you manage your entire MCP ecosystem through simple JSON configs.

What It Does

MCPX acts as a centralized orchestration layer for MCP servers. Here's what you actually get:

• Dynamic routing: Direct requests to the right MCP server based on rules you define
• Zero-code integrations: Connect tools and services via JSON configuration—no boilerplate required
• Unified API surface: Expose multiple disparate MCP services through a single, consistent interface
• Local-first with remote flexibility: Keep your workflows local while preparing for production agent deployments
• Permission enforcement: Control who (or what AI agent) can access which services and capabilities

It's built for teams orchestrating agent workflows at scale or developers building secure, multi-tool environments without maintaining separate integrations for each service.

How to Install

Pull the Docker image and run it:

docker run --rm --pull always --privileged -v ./:/lunar/packages/mcpx-server/config -p 9000:9000 -p 5173:5173 -p 3000:3000 --name mcpx us-central1-docker.pkg.dev/prj-common-442813/mcpx/mcpx:latest

Add it to your Claude Desktop configuration:

{
  "mcpServers": {
    "mcpx---mcp-gateway": {
      "command": "docker run --rm --pull always --privileged -v ./:/lunar/packages/mcpx-server/config -p 9000:9000 -p 5173:5173 -p 3000:3000 --name mcpx us-central1-docker.pkg.dev/prj-common-442813/mcpx/mcpx:latest"
    }
  }
}

Real-World Use Cases

• Multi-service agent workflows: Route a single Claude conversation through different MCP servers (API integrations, database connectors, file systems) without manual switching or config changes
• Production agent deployment: Set up permission boundaries and service access policies before deploying AI agents that need controlled access to production tools
• Centralized logging and auditing: Monitor all MCP interactions across your stack from a single control plane, essential for compliance-heavy environments

Full install guides for Claude Desktop, Cursor, Windsurf, and more at CuratedMCP.

Comments

[Dmitry Sergeev]

does this handle authentication for multiple users or is it mostly for a single local setup?