Agent Auth Protocol fixes non-deterministic AI identity

Traditional auth assumes a human-at-keyboard or deterministic software, creating a massive security gap for agents that currently inherit full user permissions.

• –Grants agents their own unique cryptographic identity instead of forcing them to impersonate human users or share service accounts
• –Capability-based authorization enables "just-in-time" granular permissions rather than broad, all-or-nothing OAuth scopes
• –Native support for service discovery and MCP (Model Context Protocol) allows agents to securely find and connect to tools at runtime
• –Lifecycle management enables auditing and revoking individual agents without affecting the primary user account or other active agents
• –Essential infrastructure for moving AI agents from experimental scripts to production-ready enterprise workers