AI Agent protocols for 6G systems
draft-stephan-agentproto-6g-usecases-00
This document is an Internet-Draft (I-D).
Anyone may submit an I-D to the IETF.
This I-D is not endorsed by the IETF and has no formal standing in the
IETF standards process.
| Document | Type | Active Internet-Draft (individual) | |
|---|---|---|---|
| Authors | Emile Stephan , Roland Schott , Diego Lopez , Xiaodong Duan , Lionel Morand | ||
| Last updated | 2026-10-12 | ||
| Replaces | draft-stephan-ai-agent-6g | ||
| RFC stream | (None) | ||
| Intended RFC status | (None) | ||
| Formats | |||
| Stream | Stream state | (No stream defined) | |
| Consensus boilerplate | Unknown | ||
| RFC Editor Note | (None) | ||
| IESG | IESG state | I-D Exists | |
| Telechat date | (None) | ||
| Responsible AD | (None) | ||
| Send notices to | (None) |
draft-stephan-agentproto-6g-usecases-00
Agent Communication Protocols E. Stephan, Ed.
Internet-Draft Orange
Intended status: Informational R. Schott
Expires: 15 April 2027 Deutsche Telekom
D. Lopez
Telefonica
X. Duan
China Mobile
L. Morand
Huawei
12 October 2026
AI Agent protocols for 6G systems
draft-stephan-agentproto-6g-usecases-00
Abstract
Communication between AI agents and between agent and tools is
expected to be pivotal in 6G systems. The 3GPP TR 22.870 outlines
various use cases and potential service requirements for AI agent
communication within 6G systems. This document provides examples of
use cases and service requirements contained in the 3GPP TR 22.870
and extrapolates possible requirements related to agent communication
protocols.
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 15 April 2027.
Copyright Notice
Copyright (c) 2026 IETF Trust and the persons identified as the
document authors. All rights reserved.
Stephan, et al. Expires 15 April 2027 [Page 1]
Internet-Draft Agent protocols for 6G October 2026
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
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 . . . . . . . . . . . . . . . . . . . . . . . . 3
2. AI Agent related use cases in the context of 6G . . . . . . . 4
2.1. General . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.2. Network Optimization and Management . . . . . . . . . . . 5
2.3. Immersive Communications . . . . . . . . . . . . . . . . 5
2.4. Hyper-Reliable Low-Latency Communications . . . . . . . . 6
2.5. Massive IoT Device Communications . . . . . . . . . . . . 6
2.6. Security and Privacy . . . . . . . . . . . . . . . . . . 6
2.7. Autonomous Systems . . . . . . . . . . . . . . . . . . . 6
2.8. AI Agent Collaboration . . . . . . . . . . . . . . . . . 7
3. Potential agent communications related requirements . . . . . 7
3.1. General . . . . . . . . . . . . . . . . . . . . . . . . . 7
3.2. Interoperability . . . . . . . . . . . . . . . . . . . . 7
3.2.1. Standardized Protocols . . . . . . . . . . . . . . . 8
3.2.2. Multimodal Data Formats . . . . . . . . . . . . . . . 8
3.2.3. Agent Identity Management . . . . . . . . . . . . . . 8
3.3. Discovery Mechanisms . . . . . . . . . . . . . . . . . . 8
3.4. Task Management . . . . . . . . . . . . . . . . . . . . . 9
3.5. Context Awareness . . . . . . . . . . . . . . . . . . . . 9
3.5.1. Contextual Understanding . . . . . . . . . . . . . . 9
3.5.2. Adaptive Communication . . . . . . . . . . . . . . . 9
3.6. Autonomy . . . . . . . . . . . . . . . . . . . . . . . . 9
3.6.1. Decision Making . . . . . . . . . . . . . . . . . . . 9
3.6.2. Self-Management . . . . . . . . . . . . . . . . . . . 9
3.7. Security . . . . . . . . . . . . . . . . . . . . . . . . 9
3.7.1. Authentication and Authorization . . . . . . . . . . 10
3.7.2. Data Protection . . . . . . . . . . . . . . . . . . . 10
3.7.3. User Consent . . . . . . . . . . . . . . . . . . . . 10
3.8. Low Latency Communication . . . . . . . . . . . . . . . . 10
3.9. Reliability . . . . . . . . . . . . . . . . . . . . . . . 10
3.9.1. Fault Tolerance . . . . . . . . . . . . . . . . . . . 10
3.9.2. Load Balancing . . . . . . . . . . . . . . . . . . . 10
3.9.3. Redundancy . . . . . . . . . . . . . . . . . . . . . 11
3.10. Flexibility . . . . . . . . . . . . . . . . . . . . . . . 11
3.10.1. Scalability . . . . . . . . . . . . . . . . . . . . 11
3.10.2. Adaptability . . . . . . . . . . . . . . . . . . . . 11
3.10.3. Extensibility . . . . . . . . . . . . . . . . . . . 11
Stephan, et al. Expires 15 April 2027 [Page 2]
Internet-Draft Agent protocols for 6G October 2026
3.11. Energy Efficiency . . . . . . . . . . . . . . . . . . . . 11
3.11.1. Optimized Communication . . . . . . . . . . . . . . 11
3.11.2. Power Management . . . . . . . . . . . . . . . . . . 11
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 12
5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 12
6. Security Considerations . . . . . . . . . . . . . . . . . . . 12
7. References . . . . . . . . . . . . . . . . . . . . . . . . . 12
7.1. Informative References . . . . . . . . . . . . . . . . . 13
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 13
1. Introduction
Since 1998, the 3rd Generation Partnership Project (3GPP) has been
fundamental in the development of standards for various generations
of mobile networks, including 3G, 4G (LTE), and 5G.
5G has revolutionized the way we connect, offering unprecedented
throughput, low latency, and the capacity to handle a vast number of
connected devices, thus driving innovation in the consumer market and
various verticals such as healthcare, automotive, industrial
automation, satellite and smart cities. Unlike traditional networks
that rely on point-to-point interfaces, the 5G core network has been
designed as a cloud-native service-based architecture with network
functions communicating each other using RESTful APIs over HTTP/2.
These network functions can be deployed and managed dynamically,
leveraging cloud technologies such as virtualization,
containerization, and microservices. This modularity allows for more
agile, scalable, and efficient network operations.
Whereas these existing standards are still being enhanced to meet the
growing demands of the telecommunications industry, the 3GPP has
already undertaken a deep exploratory work on the use cases, service
requirements and system architecture for 6G. This study phase will
be then followed by a normative work to be completed by 2030 to meet
ITU-R IMT 2030 timeline [M.2160].
6G aims to support societal advancements and to bring value to
society in the 2030s and beyond in secure, resilient, environmentally
and economically sustainable ways. In addition to new 6G services,
other considerations are needed, e.g. CAPEX/OPEX reduction,
improvement of overall 3GPP system performance, and migration from
and interworking with 5G aspects.
A study on service requirements and use cases for 6G is documented in
the 3GPP Technical Report (TR) 22.870 [TR22.870]. While at an early
stage and the document being still a work in progress, the current
content of the report already provides useful insights on the
potential foundation pillars of the new 6G system. One of them being
Stephan, et al. Expires 15 April 2027 [Page 3]
Internet-Draft Agent protocols for 6G October 2026
the Artificial Intelligence (AI) and how 6G could leverage AI and
machine learning to enhance mobile network capabilities, service
offering and user experience.
2. AI Agent related use cases in the context of 6G
2.1. General
The recommendation ITU-R Recommendation M.2160-0 [M.2160] provides a
comprehensive framework for the development of 6G technologies,
focusing on the capabilities and objectives that these technologies
should achieve around 2030 and beyond. In this document, the
integration of AI in telecommunications is poised to be a cornerstone
for the development of 6G systems. AI is considered as a
foundational element, supporting both the network infrastructure and
devices in delivering 3GPP services, often referred to as "AI for 6G
system". Additionally, mobile network capabilities are aimed to be
enhanced and optimized for supporting AI applications, termed "6G
system for AI".
AI is not a novel concept within 3GPP, which has been actively
engaged in standardizing AI and machine learning (ML) capabilities
within 5G systems. These efforts span various domains, including
management and orchestration, core networks, and next-generation
radio access networks (NG-RAN). The objective is to boost system
performance, efficiency, and the overall end-user experience. This
ongoing work is crucial not only for current 5G advancements but also
sets a solid foundation for future 6G technologies.
Beyond traditional AI/ML capabilities, 3GPP is exploring the
integration of AI agents in 6G systems. According to 3GPP TR 22.870
[TR22.870], an AI agent is defined as follows:
Al Agent:
an automated intelligent entity that achieves a specific goal
(autonomously or not) on behalf of another entity, by e.g.
interacting with its environment, acquiring contextual
information, reasoning, self-learning, decision-making, executing
tasks (independently or in collaboration with other AI Agents).
Stephan, et al. Expires 15 April 2027 [Page 4]
Internet-Draft Agent protocols for 6G October 2026
AI agents are anticipated to enhance network efficiency by
dynamically optimizing resources, predicting network conditions, and
ensuring seamless communication between services. By incorporating
large language models (LLMs), AI agents could interpret complex
requests, convert them into actionable insights, and orchestrate
advanced 3GPP services such as immersive communication, sensing, and
computing services. It is also expected that these agents would be
able to communicate, coordinate, and cooperate with other agents to
tackle tasks that a single agent would struggle with.
In this context, communication between AI agents is expected to be
pivotal in the 6G system, by enabling advanced network
functionalities to enhance the existing capabilities of 5G networks,
providing more efficient, reliable, and secure communication
services.
The 3GPP TR 22.870 [TR22.870] outlines various use cases and
potential service requirements for AI agent communication protocols
within the 6G framework. Some of these use cases are provided in the
following sections. They are only provided for information and are
subject to change till the completion of the study.
2.2. Network Optimization and Management
* AI-Driven Network Performance Assurance: AI agents in the 6G
system should be collaborate to ensure an efficient network
configuration and resource allocation to provide on-demand 3GPP
services at a given time and location area based on the authorized
third party request by intent, e.g. such as for major sporting
events or concerts in mega venues, etc. See clause 6.43 of 3GPP
TR 22.870 [TR22.870].
* Predictive Maintenance: AI agents deployed in the 6G systems
should be able to share information to predict potential network
failures and perform preventive maintenance, reducing downtime and
enhancing network reliability. See clause 11.13 of 3GPP TR 22.870
[TR22.870].
2.3. Immersive Communications
* Traffic Management: AI agents should be able to communicate with
other agents and tools to analyze and predict traffic patterns,
optimizing bandwidth allocation and ensuring high-quality service
for users, especially in densely populated areas. See clause 6.52
of 3GPP TR 22.870 [TR22.870].
Stephan, et al. Expires 15 April 2027 [Page 5]
Internet-Draft Agent protocols for 6G October 2026
* Quality of Experience (QoE) Optimization: AI agents should be able
to collaborate to monitor and adjust network parameters in real-
time, enhancing the user experience for high-bandwidth
applications like video streaming and virtual reality. See clause
6.54 of 3GPP TR 22.870 [TR22.870].
2.4. Hyper-Reliable Low-Latency Communications
* Real-Time Decision Making: AI agents should be able to facilitate
real-time decision-making processes in critical applications such
as autonomous driving, industrial automation, and remote surgery,
where ultra-low latency and high reliability are crucial. See
clause 6.3 of 3GPP TR 22.870 [TR22.870].
* Fault Detection and Recovery: AI agents should be able to quickly
detect and recover from faults, maintaining the high reliability
required for critical applications. See clause 6.53 of 3GPP TR
22.870 [TR22.870].
2.5. Massive IoT Device Communications
* Device Management: AI agents distributed in the network (terminal,
edge, core, etc.) should be able to collaborate to manage a large
number of IoT devices, optimizing their connectivity and power
consumption to extend battery life and improve network efficiency.
See clause 6.45 of 3GPP TR 22.870 [TR22.870].
* Data Analytics: AI agents should be able to collaborate to share
and analyze data from IoT devices to provide insights and support
decision-making processes in various applications, such as smart
cities, agriculture, and healthcare. See clause 6.45 of 3GPP TR
22.870 [TR22.870].
2.6. Security and Privacy
* Anomaly Detection: AI agents should be able to exchange
information about anomalous behavior detection and potential
security threats in real-time, enhancing the overall security of
the network. See clause 6.41 of 3GPP TR 22.870 [TR22.870].
* Privacy Preservation: AI agents should be able to implement and
share privacy-preserving techniques to protect user data, maintain
sensitive data within individual networks, and ensure compliance
with regulatory requirements.
2.7. Autonomous Systems
Stephan, et al. Expires 15 April 2027 [Page 6]
Internet-Draft Agent protocols for 6G October 2026
* Autonomous Vehicles: Embedded AI agents should be able to
communicate inside the same vehicle, between vehicles, between
vehicles and the network for decision making based on environment
perception, trajectory planning, and complex real-time control.
See clause 6.3 of 3GPP TR 22.870 [TR22.870].
* Industrial Automation: AI agents should be able to communicate to
optimize and control industrial processes, improving efficiency
and reducing operational costs. See clause 6.53 of 3GPP TR 22.870
[TR22.870].
2.8. AI Agent Collaboration
* AI agent communication groups: Groups of User AI agents (AI
assistant, drone, intelligent vehicle, home robot, etc.) could be
dynamically created to allow the communication between AI agents
to complete complex tasks requested by a user and/or between
multiple user AI agent groups, owned by different users, across
diverse locations ranging from local wireless networks to the wide
area networks. See clause 6.7 of 3GPP TR 22.870 [TR22.870].
* Intelligent Communication Assistants: Operators can advantageously
provide Intelligent Communication Assistant services to their
subscribers. Intelligent Communication Assistant could understand
user intention by collecting multi-modal data of the user and
execute the user instructions by invoking other AI assistants and
services provided by the 3rd party service provider. See clause
6.11 of 3GPP TR 22.870 [TR22.870].
3. Potential agent communications related requirements
3.1. General
Agent communication in the context of AI and 6G networks involves
several common requirements to ensure effective, efficient, and
secure interactions between users and agents, between agents and
between agents and tools. Here is a list of potential key
requirements derived from the illustrative use cases provided in the
previous sections. They are not yet formally approved by 3GPP and
only provided for information/discussion.
3.2. Interoperability
Stephan, et al. Expires 15 April 2027 [Page 7]
Internet-Draft Agent protocols for 6G October 2026
3.2.1. Standardized Protocols
Agents should be able to use standardized communication protocols to
ensure they can interact seamlessly across different platforms and
systems. Agents should also support a standard protocol to interact
with external data sources (e.g. user data repository) and AI-powered
tools (e.g. location services) to complete their tasks.
3.2.2. Multimodal Data Formats
Agents should be able to support multimodal data formats (e.g., text,
file, real-time audio stream, video streaming) to facilitate easy
data exchange and interpretation.
3.2.3. Agent Identity Management
To allow the communication between users and agents as well as
between agents and/or tools across platforms/domains, the system
should support secure mechanisms for identification, verification and
governance of agents/tools identities.
3.3. Discovery Mechanisms
Robust and efficient AI agent and tool discovery mechanisms to
dynamically identify and locate AI agents/tools across different
platforms and organizations. It should be possible to combine
multiple discovery mechanisms depending on specific requirements and
constraints of the application domain, e.g. broadcasting/
multicasting-based mechanisms for intra-domain discovery and use of
centralized directories for cross-domain discovery.
Currently Service-Based Architecture (SBA) defines discovery as the
process by which Network Functions (NFs) locate and select other NFs
that provide required services. This process is mediated by the
Network Repository Function (NRF), which stores and exposes NF
profiles containing service names, versions, and operational states.
The emergence of AI functions including autonomous agents, inference
services, and retrieval-augmented generation (RAG) components, intend
based exchanges, or training challenges this model. AI functions
behave as both clients and servers within this ecosystem (but in 5G,
almost every NF both exposes and consumes services too). These
entities behave as application-layer servers that expose capabilities
such as reasoning, summarization, optimization, or prediction, and as
clients when they consume telemetry, policy, or contextual
information from the network.
Stephan, et al. Expires 15 April 2027 [Page 8]
Internet-Draft Agent protocols for 6G October 2026
Supporting discovery for the various types of AI functions requires a
broader analysis to identify solution-independent requirements.
3.4. Task Management
AI agent communication should provide a robust and efficient task
management to enable seamless coordination and collaboration among
agents. Task management includes task decomposition (complex tasks
broken down into smaller sub-tasks that can be executed by individual
agents or groups of agents), task assignment (based on agent
capabilities, availability, workload), task scheduling, task
coordination among agents and task monitoring/tracking.
3.5. Context Awareness
3.5.1. Contextual Understanding
Agents should be aware of the context in which they operate,
including the state of other agents and the environment, to make
informed decisions.
3.5.2. Adaptive Communication
Agents should be able to adapt the communication based on the context
and current needs of the system.
3.6. Autonomy
3.6.1. Decision Making
Agents should be capable of making autonomous decisions based on data
analysis and reinforcement learning but also making collaborative
decisions based on dynamic interaction between agents.
3.6.2. Self-Management
Capabilities for self-management, including self-configuration, self-
optimization, and self-healing based on information received from
other agents.
3.7. Security
Stephan, et al. Expires 15 April 2027 [Page 9]
Internet-Draft Agent protocols for 6G October 2026
3.7.1. Authentication and Authorization
The agent communications should support authentication mechanisms for
agent identity verification and authorization mechanisms to grant or
deny access to specific resources based on the authenticated agent's
permissions or roles. These mechanisms should be applicable for
intra-domain and cross-domain scenarios.
3.7.2. Data Protection
The agent communications should support encryption mechanism to
protect sensitive data from unauthorized access, ensuring privacy and
confidentiality.
3.7.3. User Consent
The agent communication should provide mechanism to collect the user
consent for the secure exchange of personal sensitive data with an AI
agent (e.g. AI assistant) inside the network or 3rd-party application
domain and between AI agents.
3.8. Low Latency Communication
The agent communication should enable minimal delay data
transmission, especially critical for real-time applications like
industrial automation, robotics, or real-time gaming, to reduce lag
and improve service experience.
3.9. Reliability
3.9.1. Fault Tolerance
The agent communication should support mechanisms to detect, mitigate
and recover from communication anomalies that would undermine
collective decision making between agents. These anomalies include
message storms, communication deadlocks, protocol violations, or
content inconsistencies.
3.9.2. Load Balancing
Agent communication protocols should be able support load-balancing
among agents to prevent any single agent from becoming a bottleneck
and maintain continuous operation.
Stephan, et al. Expires 15 April 2027 [Page 10]
Internet-Draft Agent protocols for 6G October 2026
3.9.3. Redundancy
Redundancy mechanisms (e.g. active/passive redundancy, data
redundancy, georedundancy) should be implemented to ensure high-
reliability, resilience and service continuity even if one or several
agents fail.
3.10. Flexibility
3.10.1. Scalability
Agent communication protocols should be designed to accommodate
increasing network sizes and data volumes without significant
performance degradation, even during peak load events.
3.10.2. Adaptability
Agent communication protocols should be able to adjust to multiple
contexts (including heterogeneous agent capabilities, service
environment, system requirements), multiple ways of transmitting
information, including verbal, non-verbal, written, and visual forms,
used individually or in combination, as well as the location of the
AI agents (in the user terminal, in the network, in the 3rd party
service environment).
3.10.3. Extensibility
For onward compatibility, agent communication protocols should be
able to evolve and adapt to new technologies and service requirements
without requiring major overhauls.
3.11. Energy Efficiency
3.11.1. Optimized Communication
Agent communication should support mechanisms to optimize
communication between embedded AI agents to reduce energy
consumption, especially important for battery-powered devices (e.g.
IoT devices).
3.11.2. Power Management
Agent communication should support efficient power management
strategies to extend the operational life of devices embedding AI
agents.
Stephan, et al. Expires 15 April 2027 [Page 11]
Internet-Draft Agent protocols for 6G October 2026
4. Conclusions
AI agents are envisioned to enhance network efficiency by dynamically
optimizing resources, predicting network conditions, and facilitating
seamless communication between services. The incorporation of large
language models (LLMs) will enable AI agents to understand complex
requests and orchestrate advanced services, further enhancing the
capabilities of 6G networks.
AI agent communication is expected to play a crucial role in enabling
advanced network functionalities. The use cases outlined in 3GPP TR
22.870 [TR22.870] demonstrate the potential of AI agent communication
to enhance the existing capabilities of 5G networks, providing more
efficient, reliable, and secure communication services.
In summary, the integration of AI in 6G systems represents a
significant advancement in telecommunications technology. The
ongoing work by 3GPP in standardizing AI capabilities and exploring
the potential of AI agents highlights the transformative impact that
AI is expected to have on future network infrastructures and
services.
If a multi-AI agent-based system is formally adopted by 3GPP in the
scope of 6G, standard solutions will be required to support secure
and reliable communication between agents and between agents and
external tools. These solutions will be used inside the 3GPP system
but also with 3rd-party platforms. It is then required to have
solutions developed by standard organizations that would be widely
adopted by the AI development community. It is foreseen that IETF
could be the right place to develop and maintain such standard
protocols. If such standardization work is eventually endorsed by
IETF, a close coordination between IETF and 3GPP will be essential to
ensure that any AI agent communication protocols specified by IETF
will support specific functional and service requirements defined by
3GPP in the 6G context. And it is also expected that this work will
be completed in a timely manner to cope with the challenging workplan
defined by 3GPP for the development of a 6G system in the ITU-R IMT
2030 framework [M.2160].
5. IANA Considerations
This memo includes no request to IANA.
6. Security Considerations
This document should not affect the security of the Internet.
7. References
Stephan, et al. Expires 15 April 2027 [Page 12]
Internet-Draft Agent protocols for 6G October 2026
7.1. Informative References
[TR22.870] 3GPP, "3GPP TR 22.870: Study on 6G Use Cases and Service
Requirements; Stage 1 (Release 20)",
<https://www.3gpp.org/ftp/Specs/archive/22_series/22.870>.
[M.2160] ITU-R, "Recommendation ITU-R M.2160-0: Framework and
overall objectives of the future development of IMT for
2030 and beyond", <https://www.itu.int/dms_pubrec/itu-
r/rec/m/R-REC-M.2160-0-202311-I!!PDF-E.pdf>.
Authors' Addresses
Emile Stephan (editor)
Orange
2, avenue Pierre Marzin
22300 Lannion
France
Email: emile.stephan@orange.com
Roland Schott
Deutsche Telekom
Deutsche-Telekom-Allee 9
64295 Darmstadt
Germany
Email: Roland.Schott@telekom.de
Diego Lopez
Telefonica
Email: diego.r.lopez@telefonica.com
Xiaodong Duan
China Mobile
Email: duanxiaodong@chinamobile.com
Lionel Morand
Huawei
18 QUAI DU POINT DU JOUR
92100 BOULOGNE-BILLANCOURT
France
Email: lionel.morand@huawei.com
Stephan, et al. Expires 15 April 2027 [Page 13]