HOL Guard runs on your own machine and puts a policy check in front of what an AI coding agent tries to do: shell commands, secret reads, MCP server calls, and package installs. Anything outside your policy is blocked or held for your approval before it executes. Install with pipx install hol-guard && hol-guard init. Guard Cloud is an optional paid, hosted tier for teams. It is not a gateway or network firewall.
Developer ToolsOpen SourceLocal-firstFreemiumStage: Live
Developer Tools / by Michael KantorHOL Guard
Check supported AI coding-agent actions against your policy before they run. Local, open source (Apache-2.0).
Visit project ->Why it caught our attention
HOL Guard caught our attention for putting policy enforcement close to the actions AI coding agents actually take, rather than relying only on prompts or model-level safeguards. Its local-first, open-source approach and unusually explicit documentation of coverage gaps make the security boundary easier for developers to evaluate.
byAnt visited this project and made a brief functionality check before approval. What our review covers.
The human context
Developers and teams using AI coding agents who want a local policy check before agent actions execute.
Live
Keep exploring
Related projects
Developer ToolsOpen SourceFree
CairnJS
A lightweight, zero-dependency UI framework that makes building web interfaces simple enough for beginners yet powerful enough for real projects.
Developer Toolstanstartercn
Deploy your own AI-conpatible shadcn registry in one click.
WAI Tools
Subscription
Wrenda
Reverse proxy for AI search: enriched content for crawlers, pre-rendered pages, and citation tracking.