Digital News

Intelligent agents are novel intelligent systems with capabilities such as autonomous perception, memory, decision-making, interaction, and execution. They have become a key enabler of AI products and services. As China’s intelligent agent industry moves rapidly toward large-scale adoption, intelligent agents are being increasingly deployed across a broad range of industries. Interconnection and collaboration among massive intelligent agents across different systems, platforms and scenarios have emerged as a major trend for industrial development.

To fully implement the deployment requirements in the State Council’s Opinions on Further Advancing the “AI Plus” Initiative, and to standardize and guide the innovative, safe, and orderly development of the industry, the Ministry of Industry and Information Technology guided the China Electronics Standardization Institute in organizing more than 70 key enterprises to formulate the national standards series Artificial Intelligence — Agent Interconnection (Standard No.: GB/Z 185–2026). To help all market participants in the industry accurately comprehend and efficiently implement these standards, a systematic interpretation of the core contents of the standards series is provided below.

I. Background to the Introduction of the Standards Series Artificial Intelligence — Agent Interconnection (GB/Z 185-2026)

Driven by rapid advances in AI technologies, intelligent agents have emerged as core vehicles for large-scale AI adoption and industrial deployment. They have already been widely applied in areas such as enterprise operations, industrial manufacturing, urban governance, and public services, and will further extend into physical-world applications, including embodied intelligent robots.

Nevertheless, intelligent agents built by different vendors, running on disparate platforms and architectures, currently lack unified specifications covering identity identification, capability description, agent discovery & matching, interactive collaboration, and tool invocation. This leads to excessive adaptation costs, poor collaboration efficiency, and challenges in trusted management. Although protocols such as MCP and A2A have emerged internationally, no unified consensus has been formed, and such protocols fail to fully meet the needs of industry development.

The standards series Artificial Intelligence — Agent Interconnection (GB/Z 185-2026) is designed to build a unified technical framework and normative basis, standardize core links of agent interconnection, boost interoperability, composability and industrial engineering practicability, and advance standardized, large-scale, high-quality development of China’s intelligent agent industry.

II. Overall Approach and Key Contents of the Standards Series Artificial Intelligence — Agent Interconnection (GB/Z 185-2026)

Upholding the principles of systematicity, advanced technical performance and operability, this standards series establishes a standardized technical framework covering the entire lifecycle of agent interconnection, from “trusted identity, visible capabilities, discovery & matching, and interactive collaboration” to “tool invocation and task fulfillment.”

The series consists of seven parts:

Part 1:specifies the overall architecture for intelligent agent interconnection;

Part 2:specifies the encoding, assignment, and management of intelligent agent identity codes;

Part 3:specifies agent identity registration, accounts, credentials, and authentication;

Part 4:specifies the agent capability description and its registration, release, and updates;

Part 5:specifies the agent discovery process;

Part 6:specifies interaction modes for intelligent agents, including peer-to-peer, group-based, and hybrid interactions;

Part 7:specifies the architecture, processes, and data formats for external tool invocation.

Collectively, these seven parts form a closed-loop task system: identity codes and identity management mechanisms enable the identification of interconnected agents and trust verification; agent capability description and discovery mechanisms clarify what agents can do and how they can be discovered; and interaction and tool invocation mechanisms support multi-agent task collaboration and access to external resources. Together, they constitute a complete standard system for intelligent agent interconnection.

In terms of the overall architecture, the standards define five conceptual domains: the User Domain is responsible for task initiation and result reception;the Agent Domain handles identity maintenance, description maintenance, interconnection authentication, interaction, and tool access; the Management Service Domain provides identity, credential, and authentication services; the Interconnection Service Domain provides description management, agent discovery, and message routing services; and the Resource Access Domain provides access to tools and external resources. Based on a layered, decoupled architectural design, the standards effectively address key requirements for secure and controllable agent interconnection, dynamic scalability, interoperability among heterogeneous systems, and flexible evolution, embodying a forward-looking architectural design philosophy.

III. Ensuring Secure and Trustworthy Agent Interconnection through the Standards Series Artificial Intelligence — Agent Interconnection (GB/Z 185-2026)

Agent interconnection requires not only basic connectivity but also trustworthiness, controllability, and traceability. The standards series Artificial Intelligence — Agent Interconnection (GB/Z 185-2026) takes trusted identity as the cornerstone of agent interconnection. Supported by mechanisms covering identity codes, identity accounts, credentials, and identity authentication, the standards provide security assurance for cross-system multi-agent collaboration.

For identity identification, agent identity codes are centrally allocated by identity registration service providers. A hierarchical identifier system is used to support agent identification, verification, and management. Each identity code corresponds exclusively to one single agent, ensuring uniqueness and traceability from the outset.

For identity management, the standards cover the full lifecycle, including agent registration and verification, account management, credential management, and identity authentication. Once an agent passes verification, the registration service provider creates an account, allocates an identity code, and issues credentials. Subsequent operations, including update, suspension, reactivation, and revocation, are supported to realize continuous management.

For interconnection and collaboration, requesting agents and service agents shall complete mutual identity authentication before any interaction. Access control and interconnection authorization are enforced based on identity credentials and authentication results. Trusted identity verification and permission control run through the entire process, including registration, discovery, interaction, and tool invocation, laying a solid security foundation for large-scale agent interconnection.

IV. Impact of the Standards Series Artificial Intelligence — Agent Interconnection (GB/Z 185-2026) on Agent Industry Deployment, Ecosystem Development, and User Experience

The implementation of the standards series will drive intelligent agents to evolve from isolated applications and closed siloed systems to cross-platform, cross-system and cross-scenario collaborative operation, delivering fundamental underpinnings for large-scale industrial adoption.

In industrial deployment,the standards establish unified mechanisms for identity management, capability description, discovery & matching, interactive collaboration, and tool invocation. They help address challenges such as inconsistent identities, lack of visibility into agent capabilities, and incompatible interfaces, reducing costs for system integration and scenario adaptation. This speeds up the deployment of intelligent agents across enterprise services, manufacturing, government administration, and embodied intelligence sectors.

In ecosystem development,the standards deliver unified normative bases for product R&D, platform development, system integration, testing, and verification. They enable unified management, scheduling, combination, and reuse of heterogeneous intelligent agents, promote open capability sharing, and cultivate an open, collaborative industrial ecosystem.

In user experience,the standards will boost the credibility, convenience, and continuity of services. Users can clearly identify each agent’s identity, functional scope and trust level, and leverage multi-agent collaboration and tool invocation to complete complex tasks efficiently.

Overall, the standards will help lower barriers to industrial deployment, enhance enterprise collaboration efficiency, improve user experience, and foster an orderly, sound industrial ecosystem, facilitating the application of AI across a wide range of industries.

Source: Official WeChat Account of the Ministry of Industry and Information Technology (ID: gxwbwx)