ThinkInAIXYZ/clawdhome

What it solves

ClawdHome provides a secure, isolated environment for running multiple AI agent instances on a single macOS machine. It solves the problem of "weak isolation" (where browser profiles share a single user account) and "heavy isolation" (where VMs or Docker cannot run native macOS apps), allowing users to run diverse agent teams without risking their main system credentials or leaking data between instances.

How it works

The system uses a SwiftUI control panel and a privileged XPC helper daemon to manage "Shrimps"—separate macOS user accounts created for each agent instance. Each Shrimp has its own kernel-level isolation for processes, files, and Keychain entries, as well as its own managed browser-account layer. This architecture allows the project to support multiple agent engines (OpenClaw and Hermes) side-by-side while keeping their data and permissions strictly separated.

Who it’s for

• AI studios and independent builders who need to run agents for multiple clients with physically separated credentials.
• Enterprise IT and compliance teams requiring fully on-premise deployments with traceable logs and data sovereignty.
• Developers and technical creators who want a sandbox environment to test agents or a dedicated demo environment for livestreams.

Highlights

• Dual-Engine Support: Manage both OpenClaw and Hermes agents from a single interface.
• OS-Level Isolation: Uses separate macOS user accounts to ensure a compromised agent cannot access the rest of the system.
• Extensive Connectivity: Built-in support for over 13 IM channels, including WeChat, Telegram, Slack, and WhatsApp.
• Role and Skill Ecosystem: Includes a Role Market for one-click agent team deployment and a Skills Store for extending capabilities.
• Unified Management: Integrated tools for health monitoring, watchdog auto-recovery, and layered backup/restore.