Tutorials#
Each tutorial sets up one popular AI tool to run in a Mirantis Agent Platform sandbox. The tool works as it normally does for its user, while the sandbox applies the platform's controls: default-deny egress, secrets that stay outside the sandbox, and an audit record of every request. Every tutorial includes a complete sandbox spec that you can copy and adapt.
-
Run Claude Code and Codex in herdr, a terminal workspace manager, with the agents working in a sandbox that keeps running when you close the laptop.
-
Connect VS Code to a sandbox with Remote-SSH, so the Claude Code extension runs in the sandbox instead of on your laptop.
-
Run OpenClaw as a personal assistant
Run the OpenClaw assistant with its Control UI behind your platform login and its model through managed inference.
-
Make sandboxes a Claude self-hosted environment, so tasks from the Code tab of Claude Desktop run on a machine your company operates.
These tutorials depend on third-party tools that change often, and each one pins the versions it was tested with. If a step no longer matches what you see, please check the tool's own documentation for changes.