@techreport{mpsb-agntcy-messaging-02,
    number =    {draft-mpsb-agntcy-messaging-02},
    type =      {Internet-Draft},
    institution =   {Internet Engineering Task Force},
    publisher = {Internet Engineering Task Force},
    note =      {Work in Progress},
    url =       {https://datatracker.ietf.org/doc/draft-mpsb-agntcy-messaging/02/},
    author =    {Luca Muscariello and Michele Papalini and Mauro Sardara and Sam Betts},
    title =     {{An Overview of Messaging Systems and Their Applicability to Agentic AI}},
    pagetotal = 22,
    year =      2026,
    month =     jul,
    day =       18,
    abstract =  {Agentic AI systems require messaging infrastructure that supports real-time collaboration, high-volume streaming, and dynamic group coordination across distributed networks. Traditional protocols like AMQP {[}AMQP{]}, MQTT {[}MQTT{]}, and NATS {[}NATS{]} address some requirements but fall short on security, particularly regarding post-compromise protection and forward secrecy essential for autonomous agents handling sensitive data. This document analyzes six messaging protocols—AMQP, MQTT, NATS, AMQP over WebSockets, Kafka, and AGNTCY SLIM—across dimensions critical for GenAI agent systems: streaming performance, delivery guarantees, security models, and operational complexity. We examine how each protocol's design decisions impact agentic AI deployments, from lightweight edge computing scenarios to large-scale multi- organizational collaborations. AGNTCY SLIM emerges as a purpose-built solution, integrating Message Layer Security (MLS) {[}RFC9420{]} with gRPC {[}gRPC{]} over HTTP/2 {[}RFC9113{]} to provide end-to-end encryption with forward secrecy, efficient streaming, and OAuth-based authentication {[}RFC6749{]}. Unlike transport-layer security approaches, SLIM's MLS implementation ensures secure communication even through untrusted intermediaries while supporting dynamic group membership changes essential for collaborative AI agents.},
}