



Internet Engineering Task Force                              D. Liu, Ed.
Internet-Draft                                                   Alibaba
Intended status: Standards Track                            23 July 2025
Expires: 24 January 2026


                         Agent Context Protocol
                  draft-liu-agent-context-protocol-00

Abstract

   This specification defines a standard message and protocol for
   communicating Agent Context information between AI agents.

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 24 January 2026.

Copyright Notice

   Copyright (c) 2025 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
   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.







Liu                      Expires 24 January 2026                [Page 1]

Internet-Draft              Abbreviated Title                  July 2025


Table of Contents

   1.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   2
     1.1.  Requirements Language . . . . . . . . . . . . . . . . . .   2
   2.  Architecture of Agent Context Protocol  . . . . . . . . . . .   3
   3.  Agent Context Message . . . . . . . . . . . . . . . . . . . .   5
   4.  Agent Context Protocol  . . . . . . . . . . . . . . . . . . .   7
   5.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .   7
   6.  Security Considerations . . . . . . . . . . . . . . . . . . .   8
   7.  References  . . . . . . . . . . . . . . . . . . . . . . . . .   8
     7.1.  Normative References  . . . . . . . . . . . . . . . . . .   8
   Acknowledgements  . . . . . . . . . . . . . . . . . . . . . . . .   8
   Contributors  . . . . . . . . . . . . . . . . . . . . . . . . . .   8
   Author's Address  . . . . . . . . . . . . . . . . . . . . . . . .   8

1.  Introduction

   In addition to data and content, the interaction between AI agents
   should also include the communication of agent context information.
   This context information is crucial for the called AI agent to
   generate more accurate content and complete tasks that meet user
   needs while adhering to the security and policy configurations of the
   application and internal organizations.

   Considering that there may be multiple coexisting protocols between
   AI agents, and the ways in which AI agents call each other are
   diverse, it is necessary to maintain semantic consistency and
   standardization in the transmission of agent context information.
   This is critical for improving interoperability among AI agents and
   building a scalable AI agent ecosystem.

   This specification defines the semantic standard and its protocol for
   AI agents to transmit context information, independent of the agent-
   to-agent communication protocol.  Any communication protocol between
   agents can utilize this specification to implement bindings based on
   its semantics.

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.







Liu                      Expires 24 January 2026                [Page 2]

Internet-Draft              Abbreviated Title                  July 2025


2.  Architecture of Agent Context Protocol

   As shown in figure 1, Agent Context Protocol is the "semantic layer"
   of the communication protocol stack between AI agents.















































Liu                      Expires 24 January 2026                [Page 3]

Internet-Draft              Abbreviated Title                  July 2025


                                ┌──────────────────────────────────────────────────────────────────────────────────────┐
                                │                                                                                      │
                                │                                                                                      │
┌───────────────────────────────▼────────────────────────────────┐                     ┌───────────────────────────────▼────────────────────────────────┐
│                                                                │                     │                                                                │
│  ┌──────────────────────────────────────────────────────────┐  │                     │  ┌──────────────────────────────────────────────────────────┐  │
│  │                                                          │  │                     │  │                                                          │  │
│  │                                                          │  │                     │  │                                                          │  │
│  │                                                          │  │                     │  │                                                          │  │
│  │                       Agent Context                      │  │                     │  │                       Agent Context                      │  │
│  │                                                          │  │                     │  │                                                          │  │
│  │                                                          │  │                     │  │                                                          │  │
│  └──────────────────────────────────────────────────────────┘  │                     │  └──────────────────────────────────────────────────────────┘  │
│  ┌──────────────────────────────────────────────────────────┐  │                     │  ┌──────────────────────────────────────────────────────────┐  │
│  │                                                          │  │                     │  │                                                          │  │
│  │                                                          │  │                     │  │                                                          │  │
│  │                   Orchestration                          │  │                     │  │                            Orchestration                 │  │
│  │                                                          │  │                     │  │                                                          │  │
│  │                                                          │  │                     │  │                                                          │  │
│  └──────────────────────────────────────────────────────────┘  │                     │  └──────────────────────────────────────────────────────────┘  │
│  ┌────────────────┐   ┌────────────────┐   ┌────────────────┐  │                     │  ┌────────────────┐   ┌────────────────┐   ┌────────────────┐  │
│  │                │   │                │   │                │  │                     │  │                │   │                │   │                │  │
│  │                │   │                │   │                │  │                     │  │                │   │                │   │                │  │
│  │      Model     │   │      Model     │   │      Model     │  │                     │  │      Model     │   │      Model     │   │      Model     │  │
│  │                │   │                │   │                │  │                     │  │                │   │                │   │                │  │
│  │                │   │                │   │                │  │                     │  │                │   │                │   │                │  │
│  └────────────────┘   └────────────────┘   └────────────────┘  │                     │  └────────────────┘   └────────────────┘   └────────────────┘  │
│  ┌────────────────┐   ┌────────────────┐   ┌────────────────┐  │                     │  ┌────────────────┐   ┌────────────────┐   ┌────────────────┐  │
│  │                │   │                │   │                │  │                     │  │                │   │                │   │                │  │
│  │  Model Context │   │  Model Context │   │  Model Context │  │                     │  │  Model Context │   │  Model Context │   │  Model Context │  │
│  │                │   │                │   │                │  │                     │  │                │   │                │   │                │  │
│  │                │   │                │   │                │  │                     │  │                │   │                │   │                │  │
│  └────────────────┘   └────────────────┘   └────────────────┘  │                     │  └────────────────┘   └────────────────┘   └────────────────┘  │
│  ┌────┐┌────┐┌────┐   ┌────┐┌────┐┌────┐   ┌────┐┌────┐┌────┐  │                     │  ┌────┐┌────┐┌────┐   ┌────┐┌────┐┌────┐   ┌────┐┌────┐┌────┐  │
│  │    ││    ││    │   │    ││    ││    │   │    ││    ││    │  │                     │  │    ││    ││    │   │    ││    ││    │   │    ││    ││    │  │
│  │Tool││Tool││Tool│   │Tool││Tool││Tool│   │Tool││Tool││Tool│  │                     │  │Tool││Tool││Tool│   │Tool││Tool││Tool│   │Tool││Tool││Tool│  │
│  │    ││    ││    │   │    ││    ││    │   │    ││    ││    │  │                     │  │    ││    ││    │   │    ││    ││    │   │    ││    ││    │  │
│  │    ││    ││    │   │    ││    ││    │   │    ││    ││    │  │                     │  │    ││    ││    │   │    ││    ││    │   │    ││    ││    │  │
│  └────┘└────┘└────┘   └────┘└────┘└────┘   └────┘└────┘└────┘  │                     │  └────┘└────┘└────┘   └────┘└────┘└────┘   └────┘└────┘└────┘  │
│                                                                │                     │                                                                │
└────────────────────────────────────────────────────────────────┘                     └────────────────────────────────────────────────────────────────┘


                     Figure 1: Diagram [REPLACE]







Liu                      Expires 24 January 2026                [Page 4]

Internet-Draft              Abbreviated Title                  July 2025


3.  Agent Context Message

   The context of an AI Agent refers to the context of information that
   the Agent can access and utilize when performing tasks or interacting
   with users.

   Providing agent context information is critical for the invoked AI
   Agent to generate more accurate results and complete tasks that meet
   user requirements, security policies, and configurations.  Well-
   defined agent contextual information enables the AI agent to better
   understand and meet user needs in a variety of complex situations,
   thereby improving its overall performance and service quality.

   This agent context information is defined as 'Agent Context Message'
   in this specification.  Agent Context Message is passed between
   orchestration layers where the network traffic originates and is
   terminated.

   The orchestration layer should create the content of Agent Context
   Message and passed as a side channel along with agent workloads.
   Each orchestration layer can choose if it is important to include the
   Agent Context Message into the model invocation prompt, depending on
   the application.  For example, a party might want the orchestrator to
   convert the JSON format of the Agent Context Message into something
   that performs better with models context windows (e.g. Markdown).  As
   another example, the Orchestrator might take the list of rule from
   the Agent Context Message and put them as guidance to the agent when
   making ethical determinations.

   This specification uses JSON to describe the Agent Context Message.
   A variety of agent communication protocols can implement binding
   using the semantics of the Agent Context Message as specified in this
   document.

   The Agent Context Message is defined as follows:

   *  Task Context: AI agents collaborate with other agents to complete
      a specific task and need to share information about the task, such
      as its progress, goals, and to-do items.  On the one hand, this
      information is provided to the collaborating agent.  It can also
      be fed back to the user, so that the user can know the overall
      progress in a timely manner.

   *  MCP Context: The MCP Context information of the Agent includes the
      MCP service list, MCP Function list, and function descriptions.
      This information can be exchanged when a connection is initiated
      between agents and can be stored in the context of the Agent.
      When the Agent's MCP service is updated, this information can be



Liu                      Expires 24 January 2026                [Page 5]

Internet-Draft              Abbreviated Title                  July 2025


      pushed to the Agents that have established connections with it,
      updating the context information.  When a user inputs a task
      request to an Agent or one Agent calls another, the MCP Context
      information can also be provided simultaneously to the other Agent
      in order to select the appropriate MCP Function.

   *  Policy Context: The organizations that deploy the Agent can
      configure policies, such as certain rules that the Agent should
      follow, and provide these policies to local or peer Agents when
      the Agent is started.

   *  Resource Context: Resource context information pertains to the
      resources required for executing tasks and tool functions, as well
      as the resources that the system can provide.

   *  Security Context: The interaction between agents needs to consider
      security aspects such as access control, authentication, and data
      protection.  The security-related context can provide information,
      for example, the authorization token that the initiator agent has
      obtained to simplify the user authorization experience.

            {
             "agentContext": {
               "taskContext": {
                 "taskId": "T123456",
                 "taskName": "Data Analysis Project",
                 "taskProgress": "30%",
                 "goals": ["Analyze Q1 sales data", "Generate predictive models"],
                 "todoItems": [
                   {"itemId": "I001", "description": "Collect raw sales data from Q1"},
                   {"itemId": "I002", "description": "Clean and preprocess the data"}
                 ],
                 "statusUpdates": [
                   {"updateId": "U001", "timestamp": "2023-10-01T12:00:00Z", "message": "Data collection completed."}
                 ]
               },
               "mcpContext": {
                 "mcpServiceList": [
                   {"serviceName": "DataFetcher", "version": "1.0.0"},
                   {"serviceName": "DataProcessor", "version": "2.0.1"}
                 ],
                 "mcpFunctionList": [
                   {"functionName": "fetchData", "description": "Fetches raw data from a specified source."},
                   {"functionName": "processData", "description": "Cleans and processes the fetched data for analysis."}
                 ]
               },
               "policyContext": {
                 "policies": [



Liu                      Expires 24 January 2026                [Page 6]

Internet-Draft              Abbreviated Title                  July 2025


                   {"policyId": "P001", "rule": "Do not share sensitive data with external parties without explicit user consent."},
                   {"policyId": "P002", "rule": "Ensure all communications are encrypted using TLS 1.2 or higher."}
                 ]
               },
               "resourceContext": {
                 "requiredResources": [
                   {"resourceType": "CPU", "quantity": "8 cores"},
                   {"resourceType": "Memory", "quantity": "16 GB RAM"}
                 ],
                 "availableResources": [
                   {"resourceType": "Storage", "quantity": "1 TB SSD"},
                   {"resourceType": "Network", "bandwidth": "1 Gbps"}
                 ]
               },
               "securityContext": {
                 "authenticationInfo": {
                   "token": "eyJhbGciOiJIUzy***"
                 },
                 "accessControl": {
                   "role": "DataAnalyst",
                   "permissions": ["read:data", "write:reports"]
                 },
                 "dataProtection": {
                   "encryptionEnabled": true,
                   "encryptionAlgorithm": "AES-256"
                 }
               }
             }
           }

                      Figure 2: Agent Context Message

4.  Agent Context Protocol

   The Agent Context Protocol is used to exchange agent context message
   defined in section 3 between agents.  The Agent Context Protocol is
   based on the JSON format and uses HTTP as the underlying transport
   protocol.  The Agent Context Protocol support bidirectional
   communication between agents(TBD).

5.  IANA Considerations

   TBD








Liu                      Expires 24 January 2026                [Page 7]

Internet-Draft              Abbreviated Title                  July 2025


6.  Security Considerations

   The Agent Context Message should be protected by encryption and
   authentication (TBD).  Since the intent of this Agent Context Message
   is to provide soft guidance and the Agent Context Message should
   never be load bearing on security.

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/info/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/info/rfc8174>.

Acknowledgements

   The author would like to thank Jason Clinton, CISO of Anthropic for
   his constructive suggestions and review.

Contributors

   TBD

Author's Address

   Dapeng Liu (editor)
   Alibaba
   Beijing
   Beijing
   100035
   China
   Email: max.ldp@alibaba-inc.com













Liu                      Expires 24 January 2026                [Page 8]
