HOL Guard is a local-first runtime firewall for AI coding agents. It sits between an AI coding agent and the machine it is running on, and it intercepts tool calls before they execute rather than reviewing them after the fact. The tool calls it intercepts include shell commands, access to secrets, MCP (Model Context Protocol) tool calls, and package installs. Because interception happens at runtime, the check applies to what the agent actually tries to do, not only to what it says it plans to do. HOL Guard is local-first: it runs on your own machine as part of your normal development environment. Installation is a single command: pipx install hol-guard && hol-guard init. The init step sets up the local configuration so the runtime firewall is in place for subsequent agent sessions. HOL Guard is released under the Apache-2.0 license, so the source is open and can be read, audited, forked, and used commercially under the terms of that license. It is free to use. Who it is for: developers and teams who run AI coding agents with real permissions on real repositories and real machines. Agents that can run shell commands, read environment variables and credential files, install packages from public registries, and call MCP servers are useful precisely because they can act, but that same capability is the risk. A single generated command can delete files, expose a secret, or pull an unexpected dependency into a project. HOL Guard's approach is to put an interception point in front of those actions so that they are visible and can be stopped at the moment they are attempted, locally, without shipping your code or your commands to a third-party service. An important caveat, stated plainly by the maker: a scan is not a safety guarantee. HOL Guard is a control that reduces exposure by intercepting agent tool calls; it is not a proof that an agent is safe, and it should be treated as one layer among the usual practices such as least-privilege credentials, sandboxes and containers, version control, code review, and backups. In summary: HOL Guard is a free, Apache-2.0 licensed, local-first runtime firewall for AI coding agents that intercepts shell, secrets, MCP, and package-install tool calls before they execute. It is made by HOL and installed with pipx install hol-guard && hol-guard init. Nothing about running a runtime firewall removes the need to pay attention to what an agent is doing on your machine, so use HOL Guard alongside good judgement, and review any tool call the firewall surfaces before you approve it and let it run.
Scheduled projects can’t be voted on until the launch week is live
AI-friendly Markdown · structured for AI citations
About HOL Guard
Ask AI about this project
Get a quick summary or comparison from ChatGPT, Claude, Gemini, Perplexity, or Mistral using this project's public listing.
Related projects
esotericAI
PremiumesotericAI - AI-Powered Tarot Readings and Astrology for your unspoken questions
11XreplyAI
AI social media tool that posts and replies in your voice across 6 platforms.
7Broodnet
Email Infrastructure for AI Agents.
4
ToolDisk.com - Discover Best AI Tools & SaaS Solutions
ToolDisk.com - Discover Best AI Tools & SaaS Solutions
4Launch AI Jam
The only launch platform with a built-in Founder Audit. List your AI tool free, get your messaging,
3Kuberns: One-Click Agentic AI Cloud Deployment
Kuberns: One-Click Agentic AI Cloud Deployment
3