HOL Guard is an open-source, local-first runtime security layer for AI agents and automation. It sits between any AI harness and the tools it wants to run, pausing risky package installs, secret reads, shell commands, and MCP changes for review before execution.

Guard supports Codex, Claude Code, Cursor, Gemini, Hermes, OpenClaw, Pi, Kimi, Grok, and other harnesses. Teams can add Guard Cloud for shared policy, fleet visibility, approvals, and evidence while enforcement continues locally on each machine.