Coding agents

What is HOL Guard? Approve AI coding agent actions before they run (Claude Code, Codex, Cursor)

HOL Guard is a free, open-source local runtime that can pause risky AI coding-agent actions for approval and keep receipts. Guard Cloud is optional and paid.

By ·

HOL Guard is a local-first runtime security layer for AI coding agents. On supported integrations, it can allow, ask or block a tool call before the action runs, and it keeps a record of what happened. The core Guard Local runtime is open source under the Apache-2.0 license. You can install it with pipx install hol-guard, without a paid plan or a cloud account. Guard Cloud is a separate, optional paid service. Based on HOL's official documentation and pricing page, plus the hashgraph-online/hol-guard repository, checked on 2026-11-14. This is not a claim of total protection: coverage depends on the agent and the event type.

OrigoTech also has a sourced HOL Guard directory listing, with the same prices verified on the official pricing page.

What it is

HOL describes Guard as a local-first runtime security layer for AI agents. It is aimed at the moment when an agent tries to take a risky action on your machine: a destructive command, access to secrets, a software install, or moving data out. On supported integrations, Guard can allow the action, ask you first, or block it, then keep a record you can review later.

The project states that Guard Local runs on your computer, that your files never leave your machine for that local runtime, and that it can keep working with no internet connection. Guard Cloud is presented as an optional layer for synchronization, shared policy, fleet visibility and team workflows, not as the component that enforces local blocking.

The current stable release advertised on the official site on 2026-11-14 was HOL Guard v3.59.0, published on 4 November 2026. The same version was listed on PyPI.

How it works

The decision path is a policy check before a supported tool call runs:

The official site and the repository describe the same surfaces: shell commands and file access, package installs, plugins, skills and agent configuration, MCP servers and tool calls, and screening of supported prompts or tool results for prompt injection and sensitive content. Guard connects through native agent hooks, managed MCP proxies and launch integrations. If a hook is missing or disabled, the README says a managed launch can be refused, as in the Codex example.

Supported agents

According to the project's README, Guard includes adapters for Codex, Claude Code, GitHub Copilot CLI, Cursor, Cline, Gemini CLI, Grok, Hermes, Kimi Code, Pi, oh-my-pi, OpenClaw, OpenCode, Antigravity, ZCode and Devin. The official site also publishes install instructions for several of these harnesses, including Claude Code, Codex and Cursor. Paseo is described as supported through those native provider integrations.

Coverage is not uniform. HOL says enforcement depth varies by agent and event type, and it publishes a support matrix that states what is blocked, reviewed, observed or not covered. An action that is only observed, or not covered, is not stopped. That limit is why this article does not describe Guard as total protection.

Install

The documented install, from the GitHub README and HOL's get-started guide, needs Python 3.10 or newer, pipx and a matching native runtime. Check the project's platform notes, including Windows ARM64, before you install:

pipx install hol-guard
hol-guard init

The first-run wizard discovers supported agents and walks you through protection setup. The README says it asks before each setup change, including opening the dashboard, installing agent integrations and connecting optional cloud services. You can check the install with hol-guard --version and hol-guard status.

The official site also offers per-agent install commands. The package itself remains hol-guard, licensed Apache-2.0, in the public hashgraph-online/hol-guard repository.

Pricing

Prices below were read on hol.org/guard/pricing on 2026-11-14. The page also offers annual billing with two months free, but the table on that page did not display a separate annual dollar amount, so those figures are not invented here. Local protection is described as free forever on one machine, with no account and no credit card required.

Pricing
PlanPriceWhat the pricing page lists
Free (Guard Local)$0 foreverLocal runtime, approvals and receipts on one machine
Solo$4.99/monthCloud sync on 2 devices, 30-day retention, 1 GB storage
Pro$15/month5 devices, 180-day retention, 5 GB storage, real-time alerts
Team$30/seat/monthUp to 25 devices, shared policy and team workflows, 365-day retention
Enterpriseon requestIdentity, log pipeline and self-hosted or VPC review

HOL's pricing FAQ says you do not need Guard Cloud to block destructive commands: Guard Local intercepts actions and enforces policy on the machine without a cloud connection. It also says Guard Local does not upload source code or secrets; if you enable Guard Cloud, sync stores redacted decision summaries, not your source code or secret material.

You can try HOL Guard (affiliate link — OrigoTech may earn a commission). The official site is hol.org/guard. The same sourced prices appear on our HOL Guard tool page.

Limits

A few limits follow from HOL's own documentation, and they matter if you are evaluating the tool:

Sources

Key takeaway: HOL Guard is a free, open-source local runtime that can allow, ask or block supported AI coding-agent tool calls before they run, and keep receipts; Guard Cloud is optional and paid, and coverage is not total.

Spotted an error in this article? Report it · Our corrections policy

OrigoTech logo

New: the app iPhone & iPad

OrigoTech is now an iPhone and iPad app

Read the articles natively, with “New” markers on unread ones, local search and offline reading of what you have already loaded. Free, no account, iOS 17 or later; ads appear between articles.

Rekha icon: a house drawn in a single stroke

From the same author Self-promotion

An Eulerian path, drawn with your finger

Rekha, the iPhone game by OrigoTech's author: connect every dot in a single stroke without tracing the same link twice, across 500 drawings that are all solvable.

The OrigoTech newsletter

A sourced weekly digest of AI tool news.

Read next

Agents

Antigravity Dynamic Workflows: /run-workflow turns agent teams into Python scripts; Copilot drops six models on December 14, Codex CLI 0.169.0

Antigravity 2.32.0 (Nov 13) adds Dynamic Workflows, where the agent writes a short Python script that runs many subagents in parallel or in sequence, and you approve the script before it runs. GitHub put Gemini 4 Hydrogen in Copilot and will retire six Copilot models on December 14. Also: Codex CLI 0.169.0, Claude Code 2.1.325, Zed 1.28 and AirTrunk's 1.8 GW plan in India.

· 6 min read

Models

Gemini 4 Hydrogen: price, benchmarks and Antigravity's new Orchestrator; Claude Code adds Code Mode, Codex CLI 0.168.0

Google released Gemini 4 Hydrogen on November 12, a fast model for everyday coding and agents at $1.50/$7.50 per million tokens, with a free API tier and a new Orchestrator in Antigravity that pairs it with Carbon or Argon. Also: Code Mode in Claude Code 2.1.323, Codex CLI 0.168.0, the agent-built AWS Emulator and Nebius's $40-45 billion capex plan for 2027.

· 5 min read