swytchcodehq/agent-install-monitor

Agent Install Monitor keeps track of everything an AI coding agent installs on your computer. When you let an AI assistant write and run code for you, it might quietly install a Python package, pull a Docker image, clone a Git repository, or start a background service. Weeks later it can be hard to remember why any of that is there. This tool keeps a local log so you always have an answer. Right now it plugs into Hermes Agent specifically, watching the commands Hermes runs and recording things like package installs across npm, pip, cargo, and many other package managers, Docker and Podman actions, Git clones, service starts, and installer downloads. Support for other AI agents such as Claude Code is planned but not available yet. The tool is described as an observability layer, not a control layer. It does not stop the agent from installing anything, and it does not undo or roll back changes on its own. Its only job is to record what happened so a human can review it later. Everything is stored locally in a small database on your machine, and the project states plainly that nothing leaves your machine. Installation runs a script that finds the same Python environment Hermes uses, installs the monitor into it so Hermes can discover it as a plugin, and offers to add a command line tool called agent-monitor to your terminal's path. Once installed and enabled, you use Hermes as normal, then run commands like agent-monitor history or agent-monitor sessions afterward in a regular terminal to review what was installed, including exporting the results as JSON or CSV. The detection method matches command text as it runs, so the author is upfront that it is a best effort log rather than a strict security boundary. It is written in Python, requires Python 3.9 or newer, and is released under the MIT license, which is a permissive license that allows free use, including commercial use.