🛡️ Runtime Antivirus for AI Agents

As AI agents gain more autonomy and tool-calling permissions, security becomes a pressing issue. This project acts as a runtime antivirus or firewall specifically designed for AI agents. Its core goal is to intercept risky tool executions, prompt injections, and malicious MCP (Model Context Protocol) servers before they can do harm. The hard part is not just identifying vulnerabilities, but blocking them at runtime without breaking the agent's workflow. It supports common environments like Claude Code and Cursor, attempting to add an LLM-specific safety net to the DevSecOps pipeline. The direction is highly relevant right now—when AI starts automatically executing scripts and accessing secrets, running without guardrails is a significant risk.