Independent software · reviewed by offstack
HOL Guard
Free, open-source approval step for AI coding agents: review risky shell, secret, MCP and install actions on your own machine first.
About HOL Guard
HOL Guard runs on the developer's own machine next to AI coding agents such as Claude Code, Codex and Cursor. When the agent tries a risky action (a shell command, reading a file that looks like a secret, an MCP tool call or a package install), Guard pauses it and asks you to approve or block it, and keeps a local record of those decisions. Install with pipx install hol-guard && hol-guard init. Limits: it is an approval point for local agent actions, not a network gateway or firewall, and it does not stop every prompt injection.
What can it do?
- Pauses risky shell commands for approval
- Flags reads of secret-looking files
- Reviews MCP tool calls and package installs
- Local record of approve/block decisions
What does free actually mean?
The local tool is free and open source (Apache-2.0) with no time limit. Guard Cloud is an optional paid add-on for teams; it is not needed to use Guard locally. Requires Python and pipx.
Best suited to
Developers who let AI coding agents run commands on their laptop
Project details are supplied by the maker. Test compatibility with your own files and workflow. Submitted by HOL.