ACMS

ACMS is an MCP server that exposes programmatic access to Apple’s Container CLI on macOS. You can run it locally or expose it over HTTP/S so MCP clients can manage containers remotely.

How to use

You use ACMS by running the server locally or pointing an MCP client at its HTTP endpoint. Once the server is up, you connect your MCP client to the provided MCP URL to issue container-related actions such as creating or managing containers on your Mac.

Key usage patterns include starting the Apple Container Services, launching the ACMS server, and then registering the MCP client to the ACMS URL. You can then perform container operations through your MCP client as if you were issuing commands to the native container CLI.

How to install

Prerequisites are required on your macOS system to run ACMS. Ensure you have a Mac with Apple Silicon, macOS 26+ (it can run with limitations on additional hardware), Xcode 26, the Apple Containerization Framework, and Apple Container CLI installed and available in PATH.

Install ACMS by choosing one of the following: clone the repository or install via Python’s package manager.

Set up and run the Apple Container Services, then start the ACMS server. The recommended approach is to use the provided startup script, or you can start the server directly with Python.

Use the following commands to get each step running on your machine.

container system start

# Recommended: Use the startup script
./start-acms.sh

# Or start directly with custom options
python3 acms/acms.py --port 8765 --host 127.0.0.1 > acms.log 2>&1 &

# Configure MCP client
claude mcp add --transport http acms http://localhost:8765/mcp

Configuration and notes

ACMS exposes an HTTP MCP endpoint that you can consume from an MCP client. The local development URL is http://localhost:8765/mcp. Use this URL in your MCP client configuration to begin issuing container-related MCP calls.

Security considerations: running this MCP server on a remote endpoint can pose risks. Ensure you implement proper network protections and understand that exposing container management capabilities remotely increases potential data exposure.