



Network Working Group                                        M. Han, Ed.
Internet-Draft                                               Z. Han, Ed.
Intended status: Standards Track                                   T. He
Expires: 16 September 2026                                       R. Pang
                                                            China Unicom
                                                           15 March 2026


 Considerations for AI Agent Communication and Networking in Enterprise
                   draft-han-agent-comm-enterprise-00

Abstract

   This document focuses on enterprise scenarios, investigating key
   technologies including agent identification and registration,
   capability discovery, efficient communication, and secure
   collaboration.  It proposes an agent identifier and semantic routing
   mechanism to achieve trusted access and efficient collaboration among
   heterogeneous agents, providing a technical path for campus-level
   multi-agent cooperation.

Status of This Memo

   This Internet-Draft is submitted in full conformance with the
   provisions of BCP 78 and BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF).  Note that other groups may also distribute
   working documents as Internet-Drafts.  The list of current Internet-
   Drafts is at https://datatracker.ietf.org/drafts/current/.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   This Internet-Draft will expire on 16 September 2026.

Copyright Notice

   Copyright (c) 2026 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents (https://trustee.ietf.org/
   license-info) in effect on the date of publication of this document.
   Please review these documents carefully, as they describe your rights
   and restrictions with respect to this document.  Code Components



Han, et al.             Expires 16 September 2026               [Page 1]

Internet-Draft  Considerations for AI Agent Communicatio      March 2026


   extracted from this document must include Revised BSD License text as
   described in Section 4.e of the Trust Legal Provisions and are
   provided without warranty as described in the Revised BSD License.

Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   2
     1.1.  Overview  . . . . . . . . . . . . . . . . . . . . . . . .   2
   2.  Conventions and Definitions . . . . . . . . . . . . . . . . .   3
   3.  Problem Statement . . . . . . . . . . . . . . . . . . . . . .   3
     3.1.  AI Agent Protocol heterogeneity and interoperability
           barriers  . . . . . . . . . . . . . . . . . . . . . . . .   3
     3.2.  Massive Agent Connection and Addressing . . . . . . . . .   3
     3.3.  Performance, Reliability and QoS Guarantee  . . . . . . .   3
     3.4.  Trustworthy Access and Permission Control . . . . . . . .   3
   4.  Requirements of Agent Communication and Networking in
           Enterprise  . . . . . . . . . . . . . . . . . . . . . . .   4
     4.1.  Requirements for AI Agent . . . . . . . . . . . . . . . .   4
     4.2.  Requirements for Network  . . . . . . . . . . . . . . . .   4
     4.3.  Requirements for Agent Gateway  . . . . . . . . . . . . .   4
   5.  Potential Key Technologies for Network Infrastructure . . . .   4
     5.1.  Agent Service Awareness . . . . . . . . . . . . . . . . .   5
     5.2.  Agent Service Bearer  . . . . . . . . . . . . . . . . . .   5
     5.3.  Agent ID and Skill Identification . . . . . . . . . . . .   5
     5.4.  Token/Agent-aware Routing . . . . . . . . . . . . . . . .   5
   6.  Potential Key Technologies for AI Agent . . . . . . . . . . .   6
   7.  Potential Key Technologies for AI Agent Gateway . . . . . . .   6
   8.  Security Considerations . . . . . . . . . . . . . . . . . . .   6
   9.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .   6
   10. Informative References  . . . . . . . . . . . . . . . . . . .   6
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .   7

1.  Introduction

1.1.  Overview

   Artificial intelligence is evolving from a model-dominated stage
   toward autonomous and collaborative agents.  Constructing network
   infrastructure that supports efficient and secure agent collaboration
   has become critical.  To address the core challenges of large-scale
   multi-agent collaboration in enterprise digital transformation, this
   paper focuses on enterprise and campus scenarios and proposes an
   agent private network architecture centered on the Agent Gateway as
   the core device.  It investigates a key technical system covering
   agent identification and registration, capability discovery,
   efficient communication, and secure collaboration, and innovatively
   proposes an IPv6 address-based agent identifier and semantic routing
   mechanism.  Based on the Agent Gateway, this architecture achieves



Han, et al.             Expires 16 September 2026               [Page 2]

Internet-Draft  Considerations for AI Agent Communicatio      March 2026


   trusted access, capability discovery, and efficient collaboration
   among heterogeneous agents through synergy between network-layer and
   application-layer capabilities, providing a technical reference and
   practical approach for enterprise and campus-level agent
   collaboration.

2.  Conventions and Definitions

   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
   "OPTIONAL" in this document are to be interpreted as described in BCP
   14 [RFC2119] [RFC8174] when, and only when, they appear in all
   capitals, as shown here.

3.  Problem Statement

3.1.  AI Agent Protocol heterogeneity and interoperability barriers

   There are various types of devices, platforms, and applications
   within the AI agents.  Different agents vary in resources and
   supported protocols (e.g., ACP, A2A, ANP, etc.)., so adaptation and
   protocol conversion are required for agents' communication.

3.2.  Massive Agent Connection and Addressing

   The number of agents in the enterprise (including robots, sensors,
   digital humans, edge AI units, etc.) can reach the million-level
   scale.  Traditional IP addressing and service discovery mechanisms
   (such as DNS and static configuration) are unable to support dynamic
   registration, real-time discovery, and high-concurrency connections,
   which leads to low addressing accuracy and high communication
   latency.

3.3.  Performance, Reliability and QoS Guarantee

   AI agent services (such as real-time control, video analysis, and
   emergency response) have stringent requirements for low latency, high
   reliability, and deterministic QoS.  However, the best-effort
   transmission mode of traditional networks fails to meet these strict
   demands, which easily gives rise to issues such as transmission
   delay, jitter, packet loss, and congestion collapse.

3.4.  Trustworthy Access and Permission Control

   During the cross-domain or cross-system communication of agents,
   there is a lack of unified identity authentication, fine-grained
   permission control, and behavior auditing mechanisms.  This brings
   about potential security risks.



Han, et al.             Expires 16 September 2026               [Page 3]

Internet-Draft  Considerations for AI Agent Communicatio      March 2026


4.  Requirements of Agent Communication and Networking in Enterprise

4.1.  Requirements for AI Agent

   Agent Identity Identification & Discovery: Distinguish agents from
   real users, integrate semantic understanding and addressing to match
   business needs with corresponding agents.  Heterogeneous
   Compatibility: Different agents vary in resources and supported
   protocols, so adaptation and protocol conversion are required for
   agents communication.

4.2.  Requirements for Network

   Efficient Agent Communication: For agent interaction, consider
   network reachability, QoS guarantee, and differentiated SLA services.
   Due to high pressure in peer-to-peer collaboration among multiple
   agents, convergence is needed to enable hierarchical communication.
   Agent Scheduling & Management: Realize agent scheduling within and
   across enterprise campus; unify orchestration of network resources
   and agent services to align network and business.  Security &
   Privacy: Implement access management and permission control to
   enhance security.

4.3.  Requirements for Agent Gateway

   As the core device of the enterprise agent private network, the agent
   gateway integrates the reliable transmission capability of
   communication gateways and the cognitive decision-making
   characteristics of agents.  It acts as the connection center, policy
   enforcer, and security guardian, supporting trusted access, efficient
   communication, and collaborative operations among agents.

   [I-D.han-rtgwg-agent-gateway-intercomm-framework] proposes the Agent
   Gateway Intercommunication Framework, with Application Service Layer,
   Orchestration & Control Layer, Agent Connectivity Layer, Network
   Communication Layer.

5.  Potential Key Technologies for Network Infrastructure

   This section proposes several potential technologies, which are
   mainly used for agent communication and networking in enterprise
   scenarios.









Han, et al.             Expires 16 September 2026               [Page 4]

Internet-Draft  Considerations for AI Agent Communicatio      March 2026


5.1.  Agent Service Awareness

   Agent Service Type Identification: Identifies services including
   real-time control, video analysis, semantic interaction, data
   synchronization, and emergency commands.  QoS Requirement Awareness:
   Recognizes differentiated service demands for latency, reliability,
   and throughput, and automatically matches appropriate network
   scheduling policies.  Traffic Feature and Behavior Awareness:
   Perceives agent communication patterns such as request-response,
   publish-subscribe, multicast, broadcast, and continuous streaming to
   optimize forwarding paths and resource allocation.

5.2.  Agent Service Bearer

   Deterministic Forwarding and Low-Latency Transmission: Provides
   deterministic transmission with bandwidth reservation, low jitter,
   and low packet loss for real-time control and emergency agent
   services, guaranteeing SLAs for critical services.  High-Concurrency
   Connection Bearer: Supports concurrent access of massive agents, long
   connection keepalive, and lightweight signaling interaction, solving
   resource bottlenecks associated with million-scale agent access.

5.3.  Agent ID and Skill Identification

   Globally Unique Agent Identifier: Defines a unified, routable, and
   cross-domain interoperable Agent ID to replace traditional IP
   addresses as the core identity of agents.  Standardized Description
   of Capabilities/Skills: Performs structured and parsable skill
   modeling for functions provided by agents, including perception,
   reasoning, execution, collaboration, and scheduling.  ID and Skill
   Registration and Publication: Supports automatic registration of an
   agent's ID, attributes, skills, location, and status, forming a
   global directory service.

5.4.  Token/Agent-aware Routing

   Token-aware Routing: Forwards the first token and critical tokens
   with high priority to reduce first-packet latency.  Combined with
   token importance classification, differentiated QoS is achieved.
   Agent-aware Routing: Performs routing based on Agent ID, Skill, and
   intent; automatically selects the optimal service provider according
   to task objectives, capability requirements, and load status;
   supports cross-domain, cross-gateway, and cross-enterprise agent
   addressing.  It realizes routing by capability, forwarding by intent
   and supports automatic collaboration among multiple agents.






Han, et al.             Expires 16 September 2026               [Page 5]

Internet-Draft  Considerations for AI Agent Communicatio      March 2026


6.  Potential Key Technologies for AI Agent

   TBD.

7.  Potential Key Technologies for AI Agent Gateway

   TBD.

8.  Security Considerations

   TBD.

9.  IANA Considerations

   TBD.

10.  Informative References

   [I-D.han-rtgwg-agent-gateway-intercomm-framework]
              Zhengxin, H., Ruan, Z., Han, M., Yan, J., He, T., and R.
              Pang, "Agent Gateway Intercommunication Framework", Work
              in Progress, Internet-Draft, draft-han-rtgwg-agent-
              gateway-intercomm-framework-01, 30 January 2026,
              <https://datatracker.ietf.org/doc/html/draft-han-rtgwg-
              agent-gateway-intercomm-framework-01>.

   [I-D.zgsgl-dispatch-a2a-requirements-enterprise]
              Zhang, L., Geng, N., Shang, X., Gao, Q., Li, Z., and J.
              Ge, "Enhanced A2A Requirements for Agents Collobration in
              Enterprise", Work in Progress, Internet-Draft, draft-
              zgsgl-dispatch-a2a-requirements-enterprise-01, 27 November
              2025, <https://datatracker.ietf.org/doc/html/draft-zgsgl-
              dispatch-a2a-requirements-enterprise-01>.

   [I-D.pang-agents-networking-scenarios]
              Pang, R., Han, M., Liu, B., Zhang, L., Gao, Q., Geng, N.,
              Shang, X., and Z. Li, "Agents Networking Scenarios in
              Enterprise and Broadband Networks", Work in Progress,
              Internet-Draft, draft-pang-agents-networking-scenarios-00,
              6 November 2025, <https://datatracker.ietf.org/doc/html/
              draft-pang-agents-networking-scenarios-00>.










Han, et al.             Expires 16 September 2026               [Page 6]

Internet-Draft  Considerations for AI Agent Communicatio      March 2026


   [I-D.zlgsgl-rtgwg-agents-networking-framework]
              Zhang, L., Liu, B., Geng, N., Shang, X., Gao, Q., and Z.
              Li, "Agents Networking Framework for Enterprise and
              Broadband", Work in Progress, Internet-Draft, draft-
              zlgsgl-rtgwg-agents-networking-framework-00, 3 November
              2025, <https://datatracker.ietf.org/doc/html/draft-zlgsgl-
              rtgwg-agents-networking-framework-00>.

Authors' Addresses

   Mengyao Han (editor)
   China Unicom
   Beijing
   China
   Email: hanmy12@chinaunicom.cn


   Zhengxin Han (editor)
   China Unicom
   Beijing
   China
   Email: hanzx21@chinaunicom.cn


   Tao He
   China Unicom
   Beijing
   China
   Email: het21@chinaunicom.cn


   Ran Pang
   China Unicom
   Beijing
   China
   Email: pangran@chinaunicom.cn















Han, et al.             Expires 16 September 2026               [Page 7]
