Wow! Microsoft's AI Agent supports A2A and MCP protocols, ushering in the golden age of intelligent agents

Early this morning, Microsoft announced on its official website the launch of two major development platforms, Azure AI Foundry and Microsoft Copilot Studio, which support the latest Agent development protocol A2A.

This is another key move by Microsoft following its support for MCP, and it will collaborate with Google to jointly develop and expand the A2A protocol, which is significant for the agent track. With the use of A2A and MCP protocols, agents can break down barriers related to data, development models, communication interactions, and operating environments, easily constructing large-scale complex agent automation processes.

In simple terms, the current stage of agents is akin to the Warring States period, where each entity has different technologies, data formats, and development methods. A2A + MCP can rewrite this situation like Qin Shi Huang, unifying currency and measurements to promote the integration of agents from decentralization to fusion, greatly enhancing development and automation efficiency.

Regarding Microsoft's significant moves in the agent space, Microsoft CEO Satya Nadella gave high praise, stating that open protocols like A2A and MCP are key to realizing an agent network. With Copilot Studio and Azure AI Foundry beginning to support A2A, customers will be able to build agent systems that are interoperable by design.

Netizens expressed that this is a major initiative. Microsoft's entry into A2A and MCP means that the agent network is no longer just theoretical—it has now reached the practical stage. Looking forward to seeing the subsequent developments!

Standardized protocols like A2A and MCP are the pillars for achieving scalable AI collaboration.

Agents are the future, and interoperability is key.

There is no doubt that open protocols like A2A and MCP are changing the landscape of the agent network. With the support of A2A in Copilot Studio and Azure AI Foundry, customers can seamlessly create interoperable agent systems

I am pleased to see that Microsoft has adopted an open protocol to achieve interoperability among agents. A2A and MCP will become important cornerstones of the agent network.

Why Microsoft Supports A2A and MCP

According to the development data released by Microsoft, currently, more than 70,000 enterprises and digital-native companies, such as Atomicwork, Epic, Fujitsu, Gainsight, H&R Block, and LG Electronics, are developing, customizing, and managing agents and AI applications through the Azure AI Foundry development platform.

In just 4 months, over 10,000 organizations have adopted the new Agent Service to build, deploy, and scale agent systems; more than 230,000 organizations, including 90% of the Fortune 500 companies, are using Microsoft Copilot Studio to develop AI and agent applications.

Therefore, Microsoft has rich practical experience in agent development and application, and is currently one of the largest agent development platforms in the world, fully aware of the many pain points caused by the lack of interoperability among agents.

However, as more and more enterprises wish to develop complex agents to expand the scope of automation, this drawback has been magnified infinitely. A2A and MCP can effectively address this issue, helping agents achieve cross-platform, operating system, and data interoperability, simplifying the development process.

What are A2A and MCP

A2A is an interaction protocol specifically designed for agents, open-sourced by Google at the "Google Cloud Next 25" conference in April this year, with the full name "Agent 2 Agent," enabling agents to collaborate with each other, regardless of the underlying framework or vendor.

For example, a multinational manufacturing company adopts a variety of enterprise platforms and services to meet complex business needs. The SAP system is used for enterprise resource planning, efficiently integrating core business processes such as finance, supply chain, and production;

Slack serves as an important tool for internal communication and collaboration within the enterprise, allowing employees from various departments to achieve instant information transfer and project collaboration; the Oracle database is used for the storage, management, and analysis of massive production and business data.

Now, we want to automate simple operational processes of SAP, Slack, and Oracle through an agent. Previously, agents on these platforms could not communicate freely. Now, with the A2A protocol, these enterprise platforms can securely and freely automate data interactions.

Currently, more than 50 top global companies, including Microsoft, Box, Cohere, Intuit, Langchain, MongoDB, PayPal, Salesforce, SAP, ServiceNow, and UKG, have joined this protocol.

The MCP protocol was launched by the large model platform Anthropic in November last year, and its full name is "Model Context Protocol." Its main purpose is to provide a unified communication framework for large language models with external data sources, tools, and services.

MCP defines a universal format that allows AI models to schedule search engines, databases, calculators, code executors, and even other models or API services as if calling functions. Through the MCP protocol, AI applications can easily incorporate external services, functionalities, or retrieve more data, thus possessing richer capabilities.

MCP also has many development advantages: requests and returned data use JSON format, ensuring compatibility; it can seamlessly integrate with the Function Call mechanism, enabling AI to call external APIs;

It can decouple AI from business logic, allowing AI to avoid hardcoding API logic and simply choose the appropriate MCP method based on Function Call, thereby improving development efficiency.

In simple terms, MCP can be seen as the "USB interface" in the field of large models, allowing various applications to be plugged in and out without worrying about the underlying logic.

A2A Example

To promote the development of the A2A protocol, Microsoft has also participated in the development of open-source libraries. For example, due to the current lack of ready-made A2A encapsulation libraries, integration is challenging for developers. Microsoft utilized the example code from the A2A code library to integrate the Semantic Kernel agent into the A2A ecosystem, providing clear and simple integration examples.

In this example, there are several key parts. The SemanticKernelTravelManager acts like a conductor, receiving user requests and then assigning tasks to the appropriate specialized agents based on the content of the requests. For instance, if there are currency-related issues, it will be assigned to the Currency ExchangeAgent; if it involves itinerary activities, it will be assigned to the Activity PlannerAgent Currency ExchangeAgent is responsible for handling currency-related matters and integrates external tools like Frankfurter API to provide real-time exchange rates, helping users with budgeting and financial planning. Activity PlannerAgent offers personalized travel itinerary suggestions based on user preferences and budget, and can also assist in booking activities and arranging attractions.

Overall automation process: When a user submits a request to TravelManager, such as "plan a budget-friendly trip that includes currency conversion," TravelManager analyzes the needs, identifies the currency issue, and calls Currency ExchangeAgent.

Currency ExchangeAgent retrieves exchange rate information from Frankfurter API, and then Activity PlannerAgent provides suitable travel suggestions based on the budget. Finally, TravelManager compiles this information and generates a complete travel plan to return to the user.

Throughout the automation process, there are several important mechanisms. In terms of task routing and delegation, TravelManager can intelligently assign tasks to plugin-based professional agents based on context and automatic function calling capabilities. Agents can showcase their capabilities through a discovery mechanism using "Agent Card," allowing other agents to quickly find suitable partners to complete tasks.

Author of this article: AIGC Open Community, Source: AIGC Open Community, Original title: "Big News! Microsoft AI Agent Supports A2A and MCP Protocols, Entering the Golden Age of Intelligent Agents"

Risk Warning and Disclaimer

The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not take into account individual users' specific investment objectives, financial conditions, or needs. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at one's own risk