Repolode
Submit
hashgraph-online

hashgraph-online/hol-guard

Local-first runtime security layer that reviews AI agent shell commands, file access, package installs and MCP tool calls before they run.

View on GitHubWebsite
Stars
964
Forks
197
Open issues
112
Last commit
1 hour ago

What is it?

HOL Guard is an open-source protection layer for AI coding agents and their tooling. It checks actions such as shell commands, file reads, package installs and MCP tool calls before they execute, looking for secret exposure, destructive operations, prompt injection and supply-chain risks. Each decision can allow, block or request approval under a local policy, and the decision is recorded as a receipt for later review.

The CLI handles setup and everyday tasks: an init wizard discovers supported agents, `hol-guard status` and `doctor` check integrations, and commands manage approvals, receipts, an artifact inventory and package scanning. It runs without an account; an optional Guard Cloud adds shared history, team policy and fleet management. The project also ships `plugin-scanner`, a separate CLI that lints and scans agent plugins, skills and MCP packages, outputs text, JSON, Markdown or SARIF, and includes a GitHub Action.

Supported agents include Codex, Claude Code, Cursor, Gemini CLI, Cline, OpenCode and OpenClaw, among others. Coverage varies by agent, depending on which hook events each one exposes.

Key features

  • 01Reviews shell commands, file access and package installs before they run.
  • 02Blocks, allows or requests approval based on a local policy.
  • 03Records decision receipts and provides a local approval dashboard.
  • 04Scans agent plugins, skills and MCP packages with plugin-scanner.
  • 05Exports SARIF reports and provides a GitHub Action for CI gates.

Use cases

  • Protecting a developer machine while running AI coding agents such as Claude Code or Codex.
  • Gating MCP server connections and tool calls in agent workflows.
  • Checking plugin or skill packages for security issues before publishing them.
  • Adding security scans for agent plugins to a CI pipeline.

Topics

#agent-security#ai-agents#mcp-security#prompt-injection#supply-chain-security#secrets-detection#runtime-security#ai-agent-security#ai-antivirus#ai-security

FAQ

What is hol-guard?
Local-first runtime security layer that reviews AI agent shell commands, file access, package installs and MCP tool calls before they run. HOL Guard is an open-source protection layer for AI coding agents and their tooling. It checks actions such as shell commands, file reads, package installs and MCP tool calls before they execute, looking for secret exposure, destructive operations, prompt injection and supply-chain risks. Each decision can allow, block or request approval under a local policy, and the decision is recorded as a receipt for later review.
Is hol-guard free to use? What license does it have?
hol-guard is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is hol-guard still maintained?
Actively maintained. The last commit to hol-guard was on Oct 22, 2026.
What can I use hol-guard for?
Protecting a developer machine while running AI coding agents such as Claude Code or Codex. Gating MCP server connections and tool calls in agent workflows. Checking plugin or skill packages for security issues before publishing them. Adding security scans for agent plugins to a CI pipeline.
Where can I get hol-guard?
The source code and installation instructions are on GitHub: https://github.com/hashgraph-online/hol-guard

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects