HOL Guard

Apply local approval rules before AI agents take risky actions

Website

Meet HOL Guard

HOL Guard adds a local policy checkpoint to supported AI coding and automation tools. It examines covered actions before they run and can allow them, request approval, or block them according to the selected policy. Typical checks concern file deletion, access to secrets, software installation, and moving data outside the machine.

Guard integrates with tools including Claude Code, Codex, Cursor, and OpenCode, with coverage depending on the integration. Its decision path runs locally and can work offline. Activity records give users a way to review what their agent attempted and how each request was handled. The open-source project publishes release notes and setup guidance for configuring the protections around an existing workflow.

Discussion

Tried it? Share a useful thought or ask a question.

Product details

Website
hol.org
Status
Launched
Listed

Explore related topics