Microsoft’s new initiative for Agentic Web: On Sunday, May 18, 2025, Microsoft’s Chief Technology Officer, Kevin Scott, made a significant announcement at the company’s headquarters in Redmond, Washington, to journalists and analysts. He revealed that Microsoft is taking the lead in developing a technology standard to ensure collaboration among “AI agents” created by different companies.
These agents will be capable of autonomously performing specific tasks, such as fixing software bugs. Additionally, the technology will enhance the agents’ memory capabilities through “structured retrieval augmentation,” powered by the Model Context Protocol (MCP) developed by Google-backed Anthropic.
Just as the internet once transformed communication and information sharing, Microsoft aims to build an “Agentic Web” based on this concept. This platform will connect AI agents and facilitate seamless information exchange between them.
The Future of the “Agentic Web”
Kelvin Scott, the Chief Technology Officer at Microsoft, has stated that they are working on a technology that will enable AI agents to exchange information and coordinate with one another. In essence, Microsoft envisions a future “Agentic Web” where AI agents from different organizations will collaborate seamlessly, equipped with memory of past interactions.
To achieve this, they are supporting an open-source protocol called Model Context Protocol (MCP), which is maintained by a Google-affiliated company. Through MCP, AI agents will be able to access each other’s tools, data, and APIs in a standardized manner. In other words, MCP will serve as a communication bridge among AI agents.
In the future, AI agents won’t just retrieve and deliver information—they will also jointly perform complex and autonomous tasks such as data analysis, automation control, and processing various types of online orders. This collaborative execution of tasks is the core goal of the Agentic Web.
According to Scott, this concept could revolutionize the internet just like the development of the Hypertext Protocol did in the 1990s.
“In the Agentic Web, imagination—not organizational boundaries—will shape what it becomes,” Scott said.
How They Improving Memory in AI Agents?
Currently, most AI agents analyze the entire input every time to generate results; they cannot learn from past information like humans do. These AI agents have “transactional” memory, meaning they store information separately for each interaction. As a result, this leads to higher computational costs. Scott has mentioned that enhancing the memory of agents requires significant computational power; therefore, they are developing a new approach called “Structured Retrieval Augmentation.”
In fact, Microsoft wants AI agents to have memory so they can retain previous conversations. To store these conversations, they are using a method called structured retrieval augmentation, which creates a roadmap by saving concise segments of each conversation as “memory units.” This enables the AI agent, when faced with a similar task in the future, to retrieve the necessary information from its memory instead of collecting all the data from scratch every time.
“This approach aligns with the core principle of human brain training-you don’t force your brain to activate everything each time you solve a problem,” says Scott.
Importance of Model Context Protocol (MCP)
MCP, or Model Context Protocol, is an open-source protocol released by Anthropic in November 2024. By adopting MCP, Microsoft is taking the first step toward building an agentic web-enabling collaboration and communication between AI agents. This is expected to significantly enhance the efficiency and accuracy of AI agents in solving problems. Through this protocol, AI agents can collect initial information from various data sources, making them more effective and intelligent.
Additionally, based on MCP, Microsoft plans to make its AI platforms more extensible and updatable. There are also plans to integrate a variety of developer research toolkits. Integration of MCP has already begun in Copilot Studio and Azure AI Foundry. Overall, Microsoft is moving forward with a major initiative to improve the collaboration and memory capabilities of its AI agents.
According to reports, Microsoft is expected to unveil new AI development tools at its annual ‘Build’ developer conference, scheduled to be held in Seattle on May 19-20.
1 thought on “Agentic Web- Microsoft’s New Initiative in 2025: Now AI Agents Will Work Together !”