MCP Proxy
Allows installation of MCP servers directly from git repositories, enabling dynamic loading of servers from version control sources.
Supports installing MCP servers from GitHub repositories using the install_mcp_server_from_git and install_and_load_mcp_server tools.
Used as an example integration in multi-server orchestration workflows for processing payments and charges.
Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@MCP Proxyload the GitHub MCP server so I can check my pull requests"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP Proxy - Dynamic MCP Server Loading
Hot-reload MCP servers without restarting Claude Code.
One-Click Install
Cursor
VS Code
Claude Code
# Clone the repo first
git clone https://github.com/lizTheDeveloper/mcp_proxy.git
cd mcp_proxy
pip install -r requirements.txt
# Then add to Claude Code (run from parent directory of mcp_proxy)
claude mcp add mcp-proxy -- python -m mcp_proxy.servers.proxy_serverStep 1: Clone the repository
git clone https://github.com/lizTheDeveloper/mcp_proxy.git
pip install -r mcp_proxy/requirements.txtStep 2: Add to your MCP config
Claude Code (~/.claude.json or project .mcp.json):
{
"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"
}
}
}
}Cursor (~/.cursor/mcp.json):
{
"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"
}
}
}
}Note: Replace /path/to/parent/of/mcp_proxy with the actual parent directory. For example, if you cloned to /Users/me/src/mcp_proxy, use /Users/me/src.
Features
Hot-Reload: Load and reload MCP servers without restart
Dynamic Installation: Install servers from git repositories
Programmatic Orchestration: Call tools dynamically in loops and workflows
Context Savings: Load only the tools you need
Quick Start
1. Install
git clone https://github.com/lizTheDeveloper/mcp_proxy.git
cd mcp_proxy
pip install -r requirements.txt2. Configure
Add to your project's .mcp.json:
{
"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"
}
}
}
}Important: PYTHONPATH should point to the parent directory of where you cloned mcp_proxy, not the mcp_proxy directory itself. For example, if you cloned to /Users/me/src/mcp_proxy, set PYTHONPATH to /Users/me/src.
3. Use
After restarting Claude Code once to load the proxy, you can:
# Load any MCP server dynamically (no restart!)
load_mcp_server_dynamically("my-server")
# Call tools on loaded servers
call_dynamic_server_tool("my-server", "tool_name", {"param": "value"})
# Install and load from git in one step
install_and_load_mcp_server("https://github.com/user/mcp-server")Available Tools (13 total)
Tool | Description |
Dynamic Loading | |
| Load a server from .mcp.json |
| Call any tool on a loaded server |
| List currently loaded servers |
| Reload a server to pick up changes |
| Stop and unload a server |
| List all configured servers |
Installation | |
| Install from git repository |
| Install and load in one step |
| List all installed servers |
| Remove a server |
Tool Search | |
| Natural language search across loaded servers |
| List all tools from all loaded servers |
| Get detailed info about a specific tool |
How It Works
Claude Code
|
v
MCP Proxy Server (meta-tools)
|
v
Dynamic Server Loader (subprocess manager)
|
v
Individual MCP Servers (spawned on-demand)The proxy spawns MCP servers as subprocesses and communicates with them using the MCP JSON-RPC protocol over stdin/stdout.
Usage Patterns
Hot-Reload During Development
# Load your server
load_mcp_server_dynamically("my-dev-server")
# Test it
call_dynamic_server_tool("my-dev-server", "my_feature", {})
# Make code changes...
# Reload with new code
reload_mcp_server("my-dev-server")
# Test again - no restart needed!
call_dynamic_server_tool("my-dev-server", "my_feature", {})Programmatic Workflows
# Load the server
load_mcp_server_dynamically("user-management")
# Programmatic workflow
users = call_dynamic_server_tool("user-management", "list_users", {"limit": 100})
for user in users["data"]:
if user["needs_activation"]:
call_dynamic_server_tool("user-management", "activate_user", {
"user_id": user["id"]
})Multi-Server Orchestration
# Load multiple servers
for server in ["database", "stripe", "email"]:
load_mcp_server_dynamically(server)
# Orchestrate across servers
customer = call_dynamic_server_tool("database", "get_customer", {"id": 123})
payment = call_dynamic_server_tool("stripe", "charge", {"amount": 2999})
call_dynamic_server_tool("email", "send_receipt", {"to": customer["email"]})Creating Custom Servers
Create a FastMCP server and add it to .mcp.json:
# my_server.py
from fastmcp import FastMCP
mcp = FastMCP("my-server")
@mcp.tool()
def my_tool(param: str) -> dict:
"""Tool description"""
return {"success": True, "result": param}
if __name__ == "__main__":
mcp.run(){
"mcpServers": {
"my-server": {
"command": "python",
"args": ["my_server.py"],
"cwd": "/path/to/server"
}
}
}Then load it dynamically:
load_mcp_server_dynamically("my-server")
call_dynamic_server_tool("my-server", "my_tool", {"param": "hello"})License
MIT
This server cannot be installed
Resources
Unclaimed servers have limited discoverability.
Looking for Admin?
If you are the server author, to access and configure the admin panel.
Latest Blog Posts
MCP directory API
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/lizTheDeveloper/mcp_proxy'
If you have feedback or need assistance with the MCP directory API, please join our Discord server