HOL Guard
A runtime security protection tool that blocks dangerous operations before AI coding agents execute them.
About HOL Guard
HOL Guard is a security tool for AI coding agents that monitors coding agents, MCP servers, skills, plugins, and local commands, and blocks dangerous operations before they execute. It covers risky actions such as destructive commands, secret access, software installation, and data movement, and provides protection against threat categories including prompt injection, MCP security, and skill and plugin security. Its core Guard Local runtime is open source under Apache-2.0 and can be installed, inspected, and used without a paid plan or cloud account, with local protection continuing to work even when offline; Guard Cloud is an optional, separately scoped paid service for syncing evidence, sharing policies, fleet visibility, and team collaboration.
Best for
Best use cases
Using AI coding agents such as Claude Code, Codex, and OpenCode and wanting to get blocking before destructive commands, secret access, software installation, and data movement execute.
Concerned about threats such as prompt injection, inaccurate MCP tool descriptions, or hidden behavior from skill or plugin installations.
Needing continuous runtime protection in a local offline environment without relying on a cloud account.
Teams that need shared policies, dashboards, evidence export, and fleet visibility.
Compare before choosing
According to the official comparison page, AI coding agent security tools fall into three categories: runtime protection (HOL Guard), static analysis scanners (Snyk Code, GitHub Advanced Security, etc.), and more — different tools target different boundaries.
According to the official comparison page, there is no universal AI agent security platform; the right controls depend on the boundary: model traffic, MCP and tool mediation, artifacts, and more.
Features
Block dangerous AI operations
Guard blocks risky actions such as destructive commands, secret access, software installation, and data movement before they execute, keeping users in control.
Local leaked-secret scanning
Local leaked-secret scanning, runtime protection, approvals, and receipts require no account or cloud connection.
Monitor agents and MCP
Guard monitors coding agents, MCP servers, skills, plugins, and local commands.
Open source local runtime
The core Guard Local runtime is open source under Apache-2.0 and can be installed, inspected, and used without a paid plan or cloud account.
Pros
- The core Guard Local runtime is open source under Apache-2.0 and can be installed, inspected, and used without a paid plan or cloud account.
- Local leaked-secret scanning, runtime protection, approvals, and receipts require no account or cloud connection, and local protection keeps working offline.
- Blocks risky actions before they execute, covering destructive commands, secret access, software installation, and data movement.
Cons
- The Solo plan is limited to 2 devices and 30 days of history; Pro supports 5 devices; Team supports up to 25 devices.
Pricing Plans
Free
- Local protection
- Local protection modules
Solo
- 2 cloud-connected devices
- 30-day history
- Synced Guard decision history across devices
- Continuously monitors up to 5 selected GitHub repositories
- 250 credential validity checks per month
Pro
- 5 cloud-connected devices
- Longer history
- Instant alerts
- Full evidence
Team
- Up to 25 cloud-connected devices
- Shared policy and workflows
- Organization module policy local devices cannot weaken
- Shared dashboards and evidence export
* Pricing information is for reference only. Please refer to the latest announcement on HOL Guard's official website.