MakerRelay
+ Submit
← Back

HOL Guard

Approve or block risky AI coding agent actions before they execute

About

When an AI coding agent such as Claude Code, Codex or Cursor wants to run a shell command, open a secrets file, call an MCP server or install a package, HOL Guard checks the action on your machine before it runs. Routine actions pass, risky ones pause for you to allow or block, and each decision is written to a local log you can review.

Guard can also scan packages, plugins and skills before you add them. The core is Apache-2.0 and runs locally; optional paid Guard Cloud adds shared team features. Guard is not a network firewall or gateway, and it does not promise to stop every prompt injection.

Install: pipx install hol-guard && hol-guard init

Tech stack

Python

Tags

ai-agentsclaude-codedeveloper-securitymcpopen-source