Hermes Agent and Docker: setup and file access tutorial

Learn to set Hermes Agent's terminal backend to Docker, mount selected folders at /workspace, and understand which skills may run outside the container.

Player not loading? Watch on YouTube

This tutorial uses Hermes Agent Desktop to show how Docker can restrict an AI agent's terminal execution and file access. In the initial demonstration, Hermes reads a test folder containing mock personal notes, business information and credentials. The speaker notes that hiding secrets in the displayed response does not establish whether those values reached the cloud model provider. The demonstration uses GPT 5.5; it does not walk through local LLM setup.

The setup starts with Docker Desktop, which must remain running for the demonstrated workflow. The speaker sets Hermes's working directory to /workspace, adds a selected host folder under Docker volumes in config.yaml, switches the terminal backend from local to Docker, and restarts Hermes. A repeat request cannot read the unmounted test folder. Creating and copying files then demonstrates how a bind mount shares files between the container and the chosen host folder.

An Obsidian vault illustrates how to grant access to a dedicated knowledge folder. The speaker reports problems with the desktop app's automatic current-directory mount option and recommends an explicit volume entry. He also warns that some skills may execute outside the Docker terminal backend. Running the entire Hermes instance in a container through a remote gateway is discussed as a further option, with added difficulty accessing desktop apps.