On this page
HOL, short for Hashgraph Online, builds open infrastructure for AI agents. Its public products are HOL Guard, which checks risky agent actions before they run, the HOL Plugins Registry for agent plugins, skills and MCP servers, and the Registry Broker with open HCS standards for agent discovery and identity.
Most people meet HOL through one of those products first: a Guard prompt that paused a coding agent, a plugin listing with a trust score, or a Universal Agent ID in someone's agent profile. This page explains how the pieces fit, who each one is for, and where to start.
The problem HOL works on
AI agents now run shell commands, install packages, call MCP tools and talk to other agents. That raises two practical questions. Which agent, plugin or server is this, and can I trust it? And should this specific action be allowed to run on my machine?
HOL's products split along those two questions. The registry, the Registry Broker and the HCS standards handle discovery and identity. HOL Guard and the Plugin Scanner handle security, before a plugin ships and at the moment an agent tries to act. In HOL's own words, the goal is to help developers and software agents "discover services, resolve identity, exchange messages, inspect capabilities, work across registries, and add security controls at the point where automated actions execute" (About HOL).
What HOL builds
HOL Guard: runtime security for AI agents
HOL Guard sits between an AI coding agent and your computer. On supported agents such as Codex, Claude Code and Cursor, it reviews shell commands, file access, package installs and MCP tool calls before they run, then allows, blocks or asks you to approve. It runs locally without an account, and its core runtime is open source under Apache-2.0. Guard Cloud is an optional paid layer for synced history and team policy. Full explainer: What is HOL Guard?
The Plugins Registry: find agent plugins and check them first
hol.org/plugins is a public catalog of plugins, skills and MCP servers for Claude Code, Codex, Cursor, Grok, Kimi Code, OpenCode, Antigravity CLI and other agent runtimes. On October 26, 2026 it listed 50,408 plugins. Each listing shows a trust score, scanner-backed security signals and owner verification, plus install commands generated for each runtime.
HOL is explicit about the limits: a scan is not a safety guarantee, and the public catalog scanner is static scoring, not a live exploit test. Paid placements are labeled Sponsored and do not change the trust-ranked order. Listing a public GitHub repository is free.
The Registry Broker and Universal Agent IDs
The HOL Registry gives one search and one API across many agent registries and protocols, including Agentverse, Olas, Virtuals, NEAR AI, A2A and ERC-8004 agents. Every entry carries a Universal Agent ID, so the same agent can be resolved across Web2 APIs and Web3 protocols. The API lives under hol.org/registry/api/v1, with an OpenAPI document and a public MCP endpoint at hol.org/.well-known/mcp.
HCS standards
The HCS standards (HCS-1 through HCS-28) are open specifications for agent identity, profiles, messaging, registries, storage, trust scores and transparency logs. A few that show up across HOL's products:
- HCS-14: Universal Agent IDs.
- HCS-11: identity profiles.
- HCS-10: agent-to-agent messaging.
- HCS-25 and HCS-26: AI trust scores and an agent skills registry (drafts).
- HCS-28: the skill trust baseline the Plugin Scanner uses when it scores bundled skills.
The standards page says they are stewarded under Hiero through the Linux Foundation Decentralized Trust.
The Plugin Scanner
plugin-scanner is the maintainer-side tool. It checks a plugin package for leaked secrets, unsafe MCP settings, risky approval defaults and packaging problems before release, from the command line or a GitHub Action. Full explainer: What is the HOL Plugin Scanner?
Who HOL is for
- Developers who run coding agents and want a checkpoint before the agent deletes files, reads credentials or installs a package. Start with HOL Guard.
- Plugin, skill and MCP server authors who want a pre-release check and a public listing. Start with the Plugin Scanner and the Plugins Registry.
- Teams building agents that need discovery, identity or messaging across registries. Start with the Registry Broker docs and the HCS standards.
- Agents themselves. HOL publishes an OpenAPI document, an llms.txt file and an MCP endpoint so software can call it directly.
How to get started
Pick the entry point that matches the job.
Protect a coding agent on your machine
Guard needs Python 3.10 or newer. Install it with pipx, then run the guided setup, which detects your agents and asks before each change:
pipx install hol-guard
hol-guard initCheck a plugin before you publish it
pipx install plugin-scanner
plugin-scanner scan .List or find a plugin
Browse hol.org/plugins. To submit, open the Submit Plugin flow, connect GitHub and list your public repository. HOL opens a pull request on the awesome-ai-plugins list it ingests.
Build against the registry
Read the Registry API docs and the root OpenAPI document at hol.org/openapi.json. HOL also runs a credential-free sandbox at hol.org/api/agent-sandbox: a POST with a JSON body whose operation is "ping" checks your transport without billing or changing data.
Is HOL open source?
Most of it. HOL's public code lives at github.com/hashgraph-online. The HOL Guard repository, which also ships the Plugin Scanner, is Apache-2.0. The HCS specifications are published for public review. Guard Cloud and sponsored registry placements are the paid parts.
FAQ
What does HOL stand for?
HOL is short for Hashgraph Online. Hashgraph Online DAO LLC operates hol.org and maintains HOL Guard, the Plugins Registry, the Registry Broker and the HCS standards.
Is HOL the same thing as HOL Guard?
No. HOL is the organization and the platform at hol.org. HOL Guard is one of its products: the local security layer that reviews AI agent actions before they run. The Plugins Registry, the Registry Broker and the Plugin Scanner are separate products from the same team.
Is HOL free to use?
Public discovery on the registry needs no account. HOL Guard's local protection is free and open source. Guard Cloud is optional, with Solo at $4.99 a month, Pro at $15 a month and Team at $30 per seat a month (pricing). Listing a plugin is free; sponsored placements are paid and labeled.
What is the Universal Agentic Registry?
It is HOL's name for unified discovery across agent registries, MCP servers, protocols and capability indexes. In practice you reach it through the Registry Broker search and API at hol.org/registry, where each agent has a Universal Agent ID defined by HCS-14.
How can an AI agent use HOL programmatically?
Start with the OpenAPI document at hol.org/openapi.json, where every operation has an operationId and typed parameters. The public Registry MCP endpoint is at hol.org/.well-known/mcp. Protected APIs and the HOL Guard Cloud MCP endpoint use OAuth 2.0 bearer tokens with PKCE-capable flows, and HOL publishes the OAuth metadata under /.well-known.
How do I contact HOL?
Email [email protected] or use hol.org/contact. Bugs in Guard or the scanner go to the hol-guard issue tracker.
