hol-guard
hol-guard is an open-source security scanner for developers using AI coding agents, helping produce pre-execution risk checks and blocking before tools actually run.
Tool overview
Based on the available evidence, hol-guard is worth watching, but it is better treated as an early open-source security layer than a broadly validated enterprise-grade protection product. The strongest adoption evidence comes from the official GitHub repository and its own positioning: it describes itself as an AI antivirus for developer agents and explicitly lists Codex, Claude Code, Cursor, Gemini, OpenCode, plugins, skills, and MCP servers. Its 390 GitHub stars show attention and interest, but stars alone are only proof of heat, not proof that detection quality, false-positive control, or operational maturity are already strong.
In practice, this is not a code generator and not a general endpoint antivirus. A more accurate analogy is a pre-execution guardrail for AI coding agents: it adds a security check before external tools, plugins, skills, or MCP services are allowed to run, reducing the need for developers to manually inspect every tool call. The evidence supports that kind of output: an extra risk-checking step before tool execution.