HOL Guard
HOL Guard is an open-source, local-first security layer for AI agents that reviews shell commands, file access, package installs and MCP tool calls before they run, blocking or flagging secret exposure, destructive operations, prompt injection and supply-chain risks.
HOL Guard sits in front of an AI agent's actions and decides whether to allow, block or require approval for each one according to a configurable policy. It reviews shell commands and file access for destructive operations, sensitive file access and credential exposure, evaluates package-manager installs against supply-chain intelligence, inventories and reviews new or changed plugins, skills and agent configuration before they load, inspects MCP server configuration and tool calls through supported hooks and managed proxies, and screens prompts and tool results for prompt injection and sensitive content.
It runs locally without an account, with a CLI and a local dashboard for managing protection, resolving approval requests and inspecting decision history; an optional cloud service adds shared history, team policy and fleet management. It installs with pipx, and a first-run wizard discovers supported agents and walks through protection setup, asking before each change such as opening the dashboard or connecting cloud services.
It is written in Python and distributed under the Apache-2.0 license.