SuperHQ lets developers safely run multiple AI coding agents in isolated sandboxes without risking their local systems or files.

Each AI agent runs in its own microVM with writes isolated to a tmpfs overlay, keeping the host system completely protected

Supports multiple AI models including Pi, Codex, and Claude Code with native CLI/TUI interfaces

API keys are secured through a local proxy that swaps credentials on-the-wire, and users can authenticate via ChatGPT subscriptions instead

Provides a unified diff view to review and accept or discard changes made by coding agents before applying them