MCP Proxy Server

Hot-reload MCP servers on demand by loading, unloading, and orchestrating tools from multiple MCP servers.
  • python

0

GitHub Stars

python

Language

6 months ago

First Indexed

3 months ago

Catalog Refreshed

Documentation & install

Readme and setup notes from the catalogue, plus a client-ready config you can copy for your MCP host.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs
{
  "mcpServers": {
    "lizthedeveloper-mcp_proxy": {
      "command": "python",
      "args": [
        "-m",
        "mcp_proxy.servers.proxy_server"
      ],
      "env": {
        "PYTHONPATH": "/path/to/parent/of/mcp_proxy"
      }
    }
  }
}

You can hot-reload MCP servers with the MCP Proxy, allowing you to load, reload, and manage multiple MCP servers on demand without restarting the host process. This makes development and orchestration across tools faster and more flexible.

How to use

Load a server dynamically using its identifier, then call tools exposed by that server to perform tasks. You can test features, update code, and reload the server to pick up changes without restarting Claude Code. Use programmatic workflows to coordinate multiple servers and automate common tasks across your MCP ecosystem.

How to install

Prerequisites: Python and Git must be installed on your system.

Step 1: Clone the MCP Proxy repository.

git clone https://github.com/lizTheDeveloper/mcp_proxy.git
cd mcp_proxy
pip install -r requirements.txt

Step 2: Add the MCP Proxy server to your MCP configuration. Create or update your MCP config with the following entry to run the proxy server locally.

{
  "mcpServers": {
    "mcp-proxy": {
      "command": "python",
      "args": ["-m", "mcp_proxy.servers.proxy_server"],
      "cwd": "/path/to/mcp_proxy",
      "env": {
        "PYTHONPATH": "/path/to/parent/of/mcp_proxy"
      }
    }
  }
}

Additional steps you might need

After configuring, start Claude Code and load the MCP Proxy server. You can then dynamically load and manage MCP servers by name, call their tools, and perform reloading or unloading as needed.

Notes and best practices

  • Use dynamic loading to experiment with different MCP servers without disrupting your main workflow. - Keep PYTHONPATH pointing to the parent directory of mcp_proxy to ensure imports resolve correctly. - Use the provided toolset to check loaded servers, list tools, and orchestrate tasks across multiple servers.

Available tools

load_mcp_server_dynamically

Load a server from its .mcp.json configuration, enabling dynamic management without restart.

call_dynamic_server_tool

Invoke a specific tool exposed by a loaded MCP server with parameters.

get_loaded_servers

List all MCP servers currently loaded by the proxy.

reload_mcp_server

Reload a loaded server to pick up code changes without restarting the host.

unload_mcp_server

Stop and unload a previously loaded MCP server.

list_available_servers

Enumerate all configured MCP servers that can be loaded.

install_mcp_server_from_git

Install an MCP server directly from a Git repository.

install_and_load_mcp_server

Install a server from a git URL and load it in one step.

list_installed_mcp_servers

Show all MCP servers that have been installed on this system.

uninstall_mcp_server

Remove an installed MCP server from the system.

search_tools

Perform a natural-language search across tools provided by loaded servers.

list_all_tools

List all tools available from all loaded servers.

get_tool_info

Retrieve detailed information about a specific tool on a server.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational