A2A protocol explains the collaboration between agents

    I have almost stopped updating recently. First of all, I would like to say sorry to my long-term fans. So what are you doing recently? I have said before that everyone is actually doing some practices related to AI intelligent bodies. We are also doing these. During this period, we are mainly doing AI practices in data governance, including the application of knowledge graphs, the practice of MCP protocols, local application tooling, and AI form design of products. I am mainly responsible for two tasks: product management and technical management, including the construction of technical architecture and pre-research of new technologies, product design and process control, and management of product development progress. So I am quite busy. In February and March, I was basically doing some research and attempts on AI. Now that the product form has been confirmed, the related design and development work has started to get on track, and I don’t have so much energy to maintain the value output of the official account.

    However, it does not mean that we have given up on updates. Instead, we need to continuously practice and verify in the process of product development in order to output more valuable content. Just as we figured out the application of MCP, Google recently output a new protocol: A2A. In this article, let’s take a look at what A2A is, and later we will talk about the relationship between A2A and MCP.

    I won’t go into too much detail about MCP. For those who don’t know much about it, you can refer to “Agent’s MCP" data-itemshowtype="0" target="_blank" linktype="text" data-linktype="2">Understand MCP that connects the world and builds intelligent AI Agents” and “【MCP】Spring Boot integrates MCP to implement local tool calls”

    Let’s take a look at A2A

A2A Introduction

    A2A, or Agent to Agent protocol, is an open source protocol launched by Google that aims to enable communication and interoperability between AI agents. By providing a standardized way for agents to collaborate , regardless of their underlying framework or vendor, the protocol enables AI agents to securely exchange information, coordinate actions, and work across various enterprise platforms and applications.

    For example, in real-world applications, A2A enables agents to connect and collaborate on complex tasks like recruiting candidates. A user can ask their agent to find candidates that match a job listing, and the agent interacts with other professional agents through A2A to find potential candidates, schedule interviews, and conduct background checks - all within a unified interface.

A2A Application Scenarios

Enterprise Automation

In an enterprise setting, A2A enables agents to work across shared data systems and applications. For example, a supply chain planning agent can be used to coordinate inventory management, logistics, and procurement agents, even if they are built by different vendors in different frameworks. This increases autonomy and improves productivity while reducing long-term costs.

Multi-agent collaboration

The A2A protocol enables true multi-agent scenarios where agents can collaborate in their natural, unstructured patterns, even if they do not share memory, tools, and context. This goes beyond simply using one agent as a "tool" for another, and allows each agent to maintain its own capabilities when handling complex tasks.

Cross-platform integration

For business applications, A2A allows AI agents to work across the entire enterprise application ecosystem. This means that agents can access and coordinate other agents across various platforms, such as CRM systems, knowledge bases, project management tools, etc. A standardized approach to managing agents across diverse platforms and cloud environments is critical to realizing the potential of collaborative AI.

Key Features of A2A

Capability Discovery

Agents can use " Agent Cards" in JSON format to describe their capabilities. This enables client agents to identify the agent that is best suited to perform a task and communicate with remote agents using A2A. For example, a client agent may discover that another agent specializes in processing financial data and delegate the task of financial analysis to it. This is similar to global supply chain selection. When you are making a product, if you cannot realize a certain technology or component yourself, you need to look for suppliers on the Internet that can provide the corresponding capabilities, and the supplier will also publish some of its own relevant information on the Internet for customers to find .

Task Management

The communication between the client and the remote agent is task-completion oriented, and the agents work together to meet user requests. This "task" object is defined by the protocol and has a lifecycle. It can be completed immediately, or for long-running tasks, each agent can communicate with each other to keep in sync. It can be understood as a project with a project manager, who looks for the corresponding partner, and the purpose of looking for a partner is to complete this factory task. The progress of each manufacturer may affect the completion time of the final task, so the project manager is required to ensure smooth communication with each manufacturer .

cooperation

Agents can send messages to each other to convey context, responses, artifacts, or user instructions. This creates a structured way for agents to share the information they need to complete tasks. For example, one agent might provide context about user preferences, while another might return analytical results.

User Experience Consultation

Each message contains "parts", which are fully formed pieces of content, such as generated text or images. Each part has a specified content type, allowing the client and remote agent to negotiate the correct format required, and explicitly includes negotiation of user UI features, such as iframes, videos, web forms, etc. Most of us understand AI as conversations, and the most common form is text messages, but because agents can do many things, such as analyzing reports, filling out forms, recording videos, opening websites, etc., the communication content between agents will also come in many forms, which requires the user experience design to take into account the form of responses provided by each agent .

A2A Design Principles

A2A design follows five key principles:

1.  Embrace agent capabilities : A2A focuses on the era in which agents can collaborate in their natural, unstructured modes, even if they do not share memory, tools, and context.

2.  Built on existing standards : The protocol is built on existing popular standards, including HTTP, SSE, and JSON-RPC, which means it is easier to integrate with the existing IT stack that enterprises already use every day.

3.  Secure by default : A2A is designed to support enterprise-level authentication and authorization, and is consistent with OpenAPI's authentication scheme at startup.

4.  Support long-running tasks : A2A is flexible and supports a variety of scenarios, from quick tasks to deep tasks that may take hours or even days. During this process, A2A can provide users with implementation feedback, notifications, and status updates.

5.  Modality-agnostic : The agent world is not limited to text, which is why A2A supports various modalities, including audio and video streams.

Advantages of A2A

Unified Intelligent Communications

A2A provides a standardized way for agents to collaborate, eliminating the need for custom integration code between different agent frameworks. This significantly reduces development effort and enables seamless communication between agents built on different platforms.

Enterprise-grade security

The protocol is designed with security as a fundamental principle and supports enterprise-grade authentication and authorization. This ensures that agent communications maintain the security standards required by business applications.

Massive interoperability

A2A enables a true multi -agent ecosystem to emerge where specialized agents can work together to solve complex problems. This interoperability enables enterprises to leverage the best agents for specific tasks, regardless of which vendor or framework they come from.

Designed for the future

By building on existing standards and supporting a variety of modalities, A2A is designed to grow as the AI ​​landscape evolves. The protocol can support emerging agent capabilities and interaction models.