mcp-server-remote

MCP server that serves a remote MCP client, ready-to-run tools and preset configuration files.

Intended for use within private network for Cybersecurity purposes.

System:  {user} <--> mcp-client-console <--> network <--> {box} <--> mcp-server-remote <--> tools
Server:  mcp-client-console <--> {mcp SDK} <--> * <--> auth.py <--> server.py <--> tools.py

User Guide | Installation

Users:

pipx install mcp-server-remote
mcp-server-remote

Developers:

git clone https://github.com/geomux/mcp-server-remote.git
cd mcp-server-remote
python3 -m venv .venv && source .venv/bin/activate
pip install -e .
mcp-server-remote

User Guide | Configuration

First run creates a default config.toml and prints its filepath (the /config command in the terminal chat prints it too). Edit it — populate with server name, url, token:

[server]
name = "Box_1"      # Label for this machine
host = "127.0.0.1"  # IP/Domain for this machine
port = 9000         # Choose port nginx proxies to
path = "/mcp"       # Leave this alone. /mcp is default for dependencies.

User Guide | Operation

Configure the MCP server first, then confirm the MCP client (see repos below) is configured to connect to it. The server accepts the connection, performs the MCP handshake, and lists the available tools to the client.

Configure UNRESTRICTED access in the Config File for the model as desired. Setting unrestricted = true will bypass all [tools.commands] allowed commands. Example

[tools]

unrestricted = true

Only recommended within a sandbox container environment due to prompt injection/model error risk

Related / Required Repos

• mcp-client-console
• mcp-sandbox-setup
• mcp-host-configuration

Project Status

• Create MCP server repo
• Connect to MCP client locally
• Connect to MCP client remotely (over the internet)
• Connect to MCP server remotely with TLS + bearer auth as requirement
• Add additional tools to tools.py