AIP: Agent Identity Protocol for Verifiable Delegation Across MCP and A2A

Researchers introduce Agent Identity Protocol (AIP) with Invocation-Bound Capability Tokens (IBCTs) to address the lack of authentication in AI agent communications via Model Context Protocol and Agent-to-Agent protocols. The protocol achieved 100% attack rejection rate in testing with minimal performance overhead of 0.086% in real deployments.

• →A scan of 2,000 MCP servers found all lacked authentication, creating security vulnerabilities in AI agent communications.
• →AIP introduces IBCTs that combine identity verification, authorization, and provenance tracking in a single token chain.
• →The protocol operates in two modes: compact JWT for single-hop cases and chained Biscuit tokens with Datalog policies for multi-hop delegation.
• →Performance testing shows minimal overhead of 0.22ms for HTTP deployment and 2.35ms (0.086%) in multi-agent systems.
• →Adversarial evaluation demonstrated 100% attack rejection rate across 600 attempts, catching unique attack vectors other systems miss.