Use Cases and Requirements for AI Agent Policy-Aware Network
draft-zhang-rtgwg-agent-policy-aware-network-01
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) | |
|---|---|---|---|
| Author | Xiaoqiu Zhang | ||
| Last updated | 2026-07-01 | ||
| 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-zhang-rtgwg-agent-policy-aware-network-01
RTGWG X. Zhang
Internet-Draft China Mobile
Intended status: Informational 1 July 2026
Expires: 2 January 2027
Use Cases and Requirements for AI Agent Policy-Aware Network
draft-zhang-rtgwg-agent-policy-aware-network-01
Abstract
With the widespread adoption of AI Agents, traditional network
architectures can no longer meet the demand for efficient
collaboration between agents and networks. This document proposes a
new paradigm of "AI Agent Policy-Aware Network", enabling three key
transformations: from Flow-aware to Agent-aware, from QoS-based to
Policy-intent-based, and from Network-driven to Agent-network
collaborative. By defining core components such as the Agent Policy-
aware Controller and Agent Policy-Aware Device, this paradigm
establishes a dynamic mapping mechanism between Agent intents and
network policies, supporting key scenarios including autonomous
performance measurement, path optimization, SLA assurance, and secure
transmission. This document outlines the background, scenarios, use
cases and requirements of Agent Policy-aware Network.
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 2 January 2027.
Copyright Notice
Copyright (c) 2026 IETF Trust and the persons identified as the
document authors. All rights reserved.
Zhang Expires 2 January 2027 [Page 1]
Internet-Draft Agent,policy-aware,network July 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 . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1. Requirements Language . . . . . . . . . . . . . . . . . . 3
1.2. Terminology . . . . . . . . . . . . . . . . . . . . . . . 3
2. Scenarios of Agent Policy-aware Networks . . . . . . . . . . 4
2.1. Medical Emergency Rescue . . . . . . . . . . . . . . . . 4
2.2. Game Multi-Agent Collaboration . . . . . . . . . . . . . 4
2.3. Government Cross-Domain Secure Transmission . . . . . . . 4
3. Architecture Overview . . . . . . . . . . . . . . . . . . . . 5
4. Use cases . . . . . . . . . . . . . . . . . . . . . . . . . . 6
4.1. Agent Policy-aware Performance Measurement . . . . . . . 6
4.2. Agent Policy-aware Path Optimization . . . . . . . . . . 7
4.3. Agent Policy-aware SLA Level Assurance . . . . . . . . . 7
4.4. Agent Policy-aware Security . . . . . . . . . . . . . . . 8
4.5. Multi-Agent Collaborative Task Orchestration . . . . . . 8
4.6. Task-Level Metering and Tiered Charging . . . . . . . . . 9
5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 10
6. Security Considerations . . . . . . . . . . . . . . . . . . . 10
7. References . . . . . . . . . . . . . . . . . . . . . . . . . 10
7.1. Normative References . . . . . . . . . . . . . . . . . . 10
7.2. Informative References . . . . . . . . . . . . . . . . . 10
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 11
1. Introduction
AI Agents are widely used in many industries. Traditional
applications passively consume network resources without independent
decision-making, while AI Agents act as autonomous decision-making
intelligent entities that interact with networks to deliver
customized policy intents and request differentiated network
capabilities. The booming adoption of AI Agents brings stricter
requirements for network service assurance. However, existing
networks built on classic QoS, APN or ARN mechanisms cannot
comprehend Agent semantic intents, leading to four major challenges:
- AI Agents lack standardized ways to explicitly convey demands like
low latency or high security to networks; - No bidirectional
collaboration channel exists between Agents and networks, blocking
end-to-end SLA guarantees; - Static resource allocation fails to fit
Zhang Expires 2 January 2027 [Page 2]
Internet-Draft Agent,policy-aware,network July 2026
Agents’ dynamic and bursty traffic characteristics; - Rigid fixed
security policies cannot adapt to diverse cross-Agent communication
requirements.
This document proposes the Agent Policy-Aware Network paradigm to
resolve these issues, realizing three core network transformations: -
From Flow-aware to Agent-aware: Networks identify traffic by
intelligent agents rather than individual flows; - From QoS-based to
Policy-intent-based: Networks interpret and execute Agents’ high-
level business intents beyond basic bandwidth/latency metrics; - From
Network-driven to Agent-network collaborative: Enable bidirectional
information exchange and joint decision making. These shifts turn
networks from passive responders into active collaborative partners
for AI Agents. This draft specifies typical scenarios, use cases and
corresponding requirements for Agent Policy-Aware Networks.
1.1. Requirements Language
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.
1.2. Terminology
Service Agent (SA): An autonomous intelligent entity capable of
perceiving environmental information, making independent decisions,
and executing tasks to achieve specific goals, which can interact
with networks to express policy intents and obtain network resources
and service guarantees.
AI Agent Policy-Aware Network: A network paradigm that can identify,
understand, and execute the policy intents of AI Agents, realizing
dynamic mapping between Agent intents and network policies, and
supporting bidirectional collaborative decision-making between Agents
and networks.
Agent Policy-Aware Controller (APAC): A core component responsible
for parsing Agent policy intents, dynamically mapping them to network
policies, and coordinating resource allocation and policy enforcement
across the network.
Agent Policy-Aware Device (APAD): A network device that can receive
and execute policies issued by the Agent Policy-aware Controller or
Service Agent, supporting real-time policy adjustment and enforcement
based on network status and Agent requirements.
Zhang Expires 2 January 2027 [Page 3]
Internet-Draft Agent,policy-aware,network July 2026
2. Scenarios of Agent Policy-aware Networks
2.1. Medical Emergency Rescue
When a home health agent detects an elderly person fainting, it needs
to immediately initiate a medical rescue process. During the rescue,
vital sign data (such as heart rate and blood pressure) must be
transmitted to the hospital in real time through an exclusive low-
latency channel to ensure the timeliness of medical decisions. At
the same time, the channel must meet high security level requirements
to prevent data leakage and strictly prohibit cross-domain
transmission to comply with medical privacy regulations. The network
system should dynamically create and maintain this channel without
manual intervention to support efficient response within the "golden
rescue time" and achieve seamless collaboration between the Agent and
the network.
2.2. Game Multi-Agent Collaboration
In real-time game battles, player Agents need to obtain low-latency
and low-jitter network guarantees during highly dynamic operations
such as team battles and skill casting. Agents should be able to
independently negotiate path switching with network Agents,
supporting "lossless switching" requirements to avoid lag, while
dynamically adapting to changes in game load. Based on Agent policy
intents (e.g., "skill casting"), the network needs to real-time
optimize transmission paths to ensure smooth operation responses.
The entire process requires no human intervention, reflecting the
collaborative capability of Agent active decision-making and network
dynamic response, and improving the real-time and immersion of the
game experience.
2.3. Government Cross-Domain Secure Transmission
Government departments need to securely transmit sensitive medical
data (such as cross-provincial health records), requiring the network
to automatically establish an ultra-high-security channel. The
security level of the channel must be dynamically adapted based on
the trust domain to which the Agent belongs (e.g., enabling group
security mechanisms within the health system), avoiding manual
configuration of whitelists or security policies. The network should
be able to identify the identities of collaborating parties (such as
digital agents of provincial health commissions and municipal medical
insurance bureaus), and enable preconfigured security policies to
ensure compliant and efficient data transmission. At the same time,
the channel needs to support cross-domain collaboration but be
strictly limited to authorized trust scopes to meet the rigid
requirements of government data security and compliance.
Zhang Expires 2 January 2027 [Page 4]
Internet-Draft Agent,policy-aware,network July 2026
3. Architecture Overview
+-----------------+
| APA Controller |
| Network Agent |
+-----------------+
/ |
/ / \
/-------- Global intent -------/ / \
/ Complicate policy / \
/ __/ \
/-----/---\ +------------+ / \ +------'------+
| Service | Local intent | APA Device |---------------| APA Device |
| Agent |--------------| Net Agent |\ /| |
\-------- / Real-time +------/-----+ \ / +-------/-----+
policy | \ / |
| \ / |
| _\/ |
| __/ \__ |
+------\-----+ __/ \_ +-------\-----+
| APA Device |/ \| APA Device |
| |----------------| |
+------\-----+ +-------------+
\ /
\ +----------+ /
\------| Cloud |-----/
+----------+
The architecture of the Agent Policy-Aware Network comprises three
core functional components: the Agent Policy-Aware Controller (APAC),
the Agent Policy-Aware Device (APAD), and service Agents including
Business Agent, Consumer Agent, and Home Agent.
AI Agents act not only as network endpoints but also as schedulers.
They explicitly convey their intents and network assurance
requirements, and autonomously select network paths or egress nodes
accordingly.
From the network perspective, APAC and APAD collaborate to identify,
interpret, decompose, and enforce the intents submitted by AI Agents.
APAD is responsible for local intent processing, policy enforcement,
and operational status monitoring to enable low-latency local
responses. APAC performs global complex intent analysis, cross-
domain resource orchestration, policy optimization, and closed-loop
lifecycle management to deliver end-to-end global network planning.
Zhang Expires 2 January 2027 [Page 5]
Internet-Draft Agent,policy-aware,network July 2026
4. Use cases
This section illustrates some use cases for Agent Policy-aware
Networks.
4.1. Agent Policy-aware Performance Measurement
Service Agents have strict latency and jitter requirements for
specific data flows (e.g., emergency rescue video streams), and need
real-time end-to-end network quality perception to ensure business
SLA compliance.
Agent Actions:
1.Collaborate with Measurement Agents on APAD, and issue structured
measurement intents based on business SLA requirements (latency,
jitter, packet loss, reliability).
2.Receive structured measurement results fed back by APAD upon task
initialization, service handover or network quality degradation, and
extract quantitative performance characteristics of business flows
(total session traffic, packet length, traffic time-series patterns,
etc.).
3.Predict traffic and security requirements based on measured
performance characteristics, map business semantic actions to network
traffic features, and output explicit QoS requirements (low latency,
anti-jitter, bandwidth reservation, ultra-high security, etc.) to the
network.
Network Actions:
1.Measurement Agents on APADs automatically select adaptive
measurement protocols (e.g., in-band flow detection, TWAMP) according
to Agent measurement intents.
2.APADs structure and feed back collected performance data to Service
Agents in a standardized format.
3.Generate unique Agent policy-aware identifiers for business flows
after receiving QoS requirements, and allocate dedicated network
resources to implement differentiated quality guarantees.
Zhang Expires 2 January 2027 [Page 6]
Internet-Draft Agent,policy-aware,network July 2026
4.2. Agent Policy-aware Path Optimization
When congestion or faults occur on the current transmission path,
service SLAs may be degraded. The network should switch to a more
optimal path without service interruption, enabling fault self-
healing and experience self-optimization, thereby ensuring continuous
service delivery.
Agent Actions:
Service Agents receive network feedback of quality degradation
alarms, and issue a request for "path recalculation and switching" to
the network in combination with service tolerance (e.g., "temporary
interruption allowed" or "lossless switching mandatory").
Network Actions:
The Agent Policy-aware Controller (APAC) calculates an optimal set of
schedulable paths that meet constraints based on real-time network
topology and link status, confirms the switching strategy with
Service Agents, and finally executes seamless path migration.
Schedulable Path Set Information Model: ‐Basic Identifier: Path ID /
Path Group ID ‐Performance Attributes: Latency, jitter, packet loss,
bandwidth, remaining bandwidth ‐O&M Attributes: Cost, priority,
reliability, congestion probability ‐Constraint Attributes: Slicing
support, cross-domain support, encryption support
4.3. Agent Policy-aware SLA Level Assurance
Different types of traffic from Agents feature high dynamism,
burstiness and periodicity (e.g., scheduled big data
synchronization). The network needs to allocate exclusive resources
such as bandwidth on demand to realize "tidal scheduling" of network
resources, greatly improve resource utilization, and ensure
deterministic experience for critical services.
Agent Actions:
Service Agents predict future traffic trends and send "resource
reservation intents" (including time window, bandwidth peak, and
duration) to the network.
Network Actions:
Zhang Expires 2 January 2027 [Page 7]
Internet-Draft Agent,policy-aware,network July 2026
The Resource Agent of APAC evaluates the global network load: if
resources are sufficient, it temporarily locks the corresponding
bandwidth slice; if resources are tight, it negotiates a degradation
scheme with the Agent or suggests adjusting the time window.
The Path Agent of APAC completes the full lifecycle management of
bearer tunnels such as SRv6 Policy and network slicing, adjusting the
level of SLA assurance and dismantling tunnels on demand.
For Agent collaboration requirements that transmit information in a
specified sequence, SRv6 service function chain evolution technology
is adopted to ensure the order and reliability of information
transmission, meeting the timing requirements of collaborative tasks.
4.4. Agent Policy-aware Security
When Agents in different trust domains conduct data interaction, the
network needs to dynamically match the corresponding security levels
and encryption mechanisms to realize identity and context-based
dynamic zero-trust security, ensuring that data only flows within
authorized scopes.
Agent Actions:
The communicating Agents exchange identity credentials and trust
domain attributes, and declare the required security levels (e.g.,
"national cryptographic algorithm encryption", "cross-domain
isolation").
Network Actions:
The Network Security Agent parses the identities of both parties,
automatically retrieves and matches preconfigured cross-domain mutual
trust policies, and provides hierarchical security technologies on
demand, including IPSec, QKD key distribution, and group secure
transmission. The policies are automatically revoked after the
session ends.
4.5. Multi-Agent Collaborative Task Orchestration
Complex service guarantee intents submitted by service agents in
intelligent operations and real-time interactive scenarios cannot be
fulfilled by a single agent alone. This use case defines mechanisms
for an orchestration agent to decompose high-level service intents
and coordinate home gateway sub-agents（for scenarios involving home
agents）, resource sub-agents, and routing sub-agents to perform joint
optimization across access and backbone network, enabling end-to-end
network guarantees via multi-agent collaboration.
Zhang Expires 2 January 2027 [Page 8]
Internet-Draft Agent,policy-aware,network July 2026
1. Experience-aware agents or service agents generates structured
service guarantee intents tagged with a globally unique task
identifier, and forwards these intents to the orchestration
agent. The orchestration agent parses service requirements and
decomposes the composite task into three discrete subtasks:
access scheduling, resource reservation, and path optimization.
Each subtask is dispatched to the corresponding dedicated sub-
agent.
2. The home gateway agent, resource agent, and routing agent submit
their respective policy requests to the APAD through the APAC.
In cases of resource shortage, cross-agent negotiation is
mediated by the orchestration agent.
3. APAD correlates all relevant service flows under a unified task
ID to support centralized task lifecycle management. Home
gateway sub-agent：adjust forwarding priority for target service
flows. Resource sub-agent：provision isolated bandwidth resources
on demand. Routing sub-agent： dynamically re-route traffic onto
optimal transmission paths.
4. APAC and APAD automatically releases all temporary scheduling
resources once the task completes. Upon successful deployment of
all subtask policies, the orchestration agent persists full
collaborative task metadata to enable subsequent integration with
metering and charging workflows.
4.6. Task-Level Metering and Tiered Charging
Temporary workloads driven by autonomous agents consume heterogeneous
network resources, creating a need for task-granular resource
accounting and differentiated charging. This use case specifies how
the network collects full resource metrics bound to task identifiers,
exposes standardized metering data via defined interfaces to a
charging agent, and facilitates tiered settlement based on task
priority and execution complexity.
1. After all task policies are provisioned, the orchestration agent
synchronizes structured task metadata to the charging agent,
including the task ID, SLA tier, task lifetime, types of
participating sub-agents, and allocated bandwidth and other
network resources.
2. The APAD continuously collects per-task resource consumption
metrics keyed by task ID, including bandwidth utilization,
isolated tunnel occupancy, and link load statistics.
Zhang Expires 2 January 2027 [Page 9]
Internet-Draft Agent,policy-aware,network July 2026
3. The APAD packages raw consumption metrics into standardized
metering records and tags all metering records with the
associated SLA tier to differentiate resource overhead incurred
by services of varying priority levels. And the APAD exposes
machine-readable interfaces for external consumption of metering
data.
4. The charging agent continuously retrieves real-time metering
statistics via the interfaces provided by the APAD.
5. Upon task termination, the charging agent aggregates full-
lifecycle resource consumption metrics and generates tiered
charging records following predefined pricing rules.
5. IANA Considerations
TBD.
6. Security Considerations
TBD.
7. References
7.1. Normative References
[RFC2119] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", BCP 14, RFC 2119,
DOI 10.17487/RFC2119, March 1997,
<https://www.rfc-editor.org/rfc/rfc2119>.
[RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
May 2017, <https://www.rfc-editor.org/rfc/rfc8174>.
7.2. Informative References
[I-D.yang-rtgwg-arn-framework-04]
Yang, F. and C. Lin, "Application-Responsive Network
Framework", Work in Progress, Internet-Draft, draft-yang-
rtgwg-arn-framework-04, 29 June 2025,
<https://datatracker.ietf.org/doc/html/draft-yang-rtgwg-
arn-framework-04>.
[I-D.li-rtgwg-apn-framework]
Li, Z., Voyer, D., Li, C., Liu, P., Cao, C., Mishra, G.
S., and N. Geng, "Application-aware Networking (APN)
Framework", Work in Progress, Internet-Draft, draft-li-
Zhang Expires 2 January 2027 [Page 10]
Internet-Draft Agent,policy-aware,network July 2026
rtgwg-apn-framework-01, 12 November 2025,
<https://datatracker.ietf.org/doc/html/draft-li-rtgwg-apn-
framework-01>.
Author's Address
Xiaoqiu Zhang
China Mobile
China
Email: zhangxiaoqiu@chinamobile.com
Zhang Expires 2 January 2027 [Page 11]