Korven

The Agent Identity Protocol (AIP) adds a security layer for autonomous AI agents by requiring cryptographic passports and intent verification before any action is performed. Each agent receives an Ed25519‑based decentralized identifier, and every request is checked against scoped permissions, monetary limits, and an allowlist of approved actions. The system supports real‑time revocation and an emergency kill switch that can instantly disable a compromised agent.

AIP is packaged as a lightweight SDK that integrates with existing AI frameworks through a simple decorator or CLI workflow. Developers can protect agents with a single line of code, defining allowed actions and usage limits, while the protocol handles boundary enforcement, tiered verification, and error reporting. First‑class adapters are provided for popular toolkits such as LangChain, CrewAI, and AutoGen, enabling seamless verification of tool calls, task execution, and multi‑agent conversations.

The protocol is open source and designed for developers who ship AI‑driven services, offering a real‑time dashboard, analytics, and webhook hooks for monitoring agent activity. By combining cryptographic identity, scoped API keys, and instantaneous revocation, it aims to prevent rogue behavior such as unauthorized spending, mass emailing, or impersonation without requiring changes to the underlying application stack.