HOL Guard – Local-First Runtime Firewall for AI Coding Agents
Overview
HOL Guard is an open-source, local-first runtime firewall for Windows that mediates risky actions by AI coding agents—including shell execution, secret access, MCP server modifications, and plugin installations.
Key Features
- Local-First Runtime Mediation: Operates entirely on-premises without routing code, credentials, or commands through external cloud gateways.
- Granular Action Control: Intercepts and validates shell commands, file reads, and .env secret access before execution.
- MCP & Plugin Defense: Prevents unauthorized Model Context Protocol (MCP) server modifications and untrusted skill installs.
- Interactive Approvals: Review explicit command payloads and choose to approve, deny, or whitelist operations.
- Open-Source Security: Released under the permissive Apache-2.0 license for transparent and auditable desktop protection.
How to Protect Your Environment
- Install HOL Guard on your Windows machine using the official MSI installer or via PyPI.
- Configure security boundary rules and permission policies for your AI coding assistants.
- Run your AI coding agents (Claude Code, Cursor, Codex, etc.) normally in your workspace.
- Review and approve or deny intercepted system calls and sensitive file requests in real time.
