This repository documents the deployment of a hardware-isolated AI agent designed for secure, real-time task automation.
- Hardware: Raspberry Pi 4 (64-bit Architecture)
- Engine: OpenClaw v2026.3.8
- LLM Provider: Anthropic Claude (via API)
- Search Layer: Brave Search API
- Communication: Telegram Bot API (Encrypted Gateway)
By hosting the agent on a dedicated Linux micro-server, I have achieved:
- Host Isolation: Prevents AI shell execution from accessing my local laptop environment.
- Credential Safety: API keys and session memories are stored in a dedicated, encrypted daemon.
- Approval Gates: All high-privilege actions require a manual "Approve" via the Telegram pairing hook.
- Integration of Local LLMs (Ollama) for offline privacy.
- AI-Driven Pentesting Skills for network vulnerability scanning.
- IoT integration for Drone/Robotics telemetry.