Aegis

Governs AI agent actions with YAML-based policies, approval gates, audit trails, and risk classification.

Aegis is a Python middleware that sits between AI agents and their actions, providing policy-based governance. It evaluates every action against YAML-defined rules with glob matching, enforces four-tier risk classification (low/medium/high/critical), and gates execution through configurable approval handlers including CLI, Slack, Discord, Telegram, and webhooks. The MCP adapter wraps Model Context Protocol tool calls with the same governance pipeline. Additional capabilities include behavioral anomaly detection, compliance reporting, cryptographic audit chains, RBAC, and multi-tenant isolation.