Microsoft is reportedly preparing to implement Anthropic’s Model Context Protocol (MCP) to enhance collaboration among multiple artificial intelligence (AI) agents. According to sources, the company based in Redmond is directing its attention towards developing multi-agent workflows alongside improved memory capabilities for its AI systems. The MCP standardizes a method for AI chatbots to engage with external data sources, facilitating the retrieval of information. An official announcement regarding the adoption of Anthropic’s MCP could come during the Microsoft Build 2025 conference set for Monday. This event primarily targets developers who are creating applications and tools within the Microsoft ecosystem.
Advancing Towards an Agentic Web
Microsoft is taking steps to standardize the interaction between its AI chatbots and external data hubs. According to the company’s Chief Technology Officer, Kevin Scott, Reuters reports that Microsoft has a dual focus for its AI agents: fostering cooperation among multiple agents and enhancing their long-term memory capabilities.
Anthropic unveiled the MCP, a universal open standard, in November 2024. This protocol addresses the challenge of linking AI systems to third-party data sources, including platforms like Google Cloud, Azure, and various cloud-based customer relationship management (CRM) solutions.
Given the diverse operational frameworks of data hubs, the processes for connecting and obtaining data can vary significantly. As a result, AI systems employing different protocols experience inconsistencies in data retrieval and processing, leading to variations in inference time and final outputs. MCP aims to establish a singular protocol for AI systems to connect with these knowledge sources and access vital data.
Scott emphasized Microsoft’s commitment to cultivating industry-wide standards that allow AI agents from various developers to collaborate. He asserted that MCP possesses the potential to foster an “agentic web,” akin to the way hypertext protocols in the 1990s facilitated the expansion of the Internet.
Additionally, Scott pointed out the necessity of enhancing the memory functions of AI agents to increase their utility. Current publicly available AI agents typically possess enough memory to handle tasks within a single prompt. Although they can generate comprehensive responses and perform online tasks, they often lack the necessary context to connect one task to the next.
To tackle this issue, Microsoft is reportedly developing a new strategy called structured retrieval augmentation. This innovative approach enables an AI agent to extract snippets of information from each interaction with a user, thereby accumulating context across a series of tasks performed during a session. Detailed information about this strategy and the integration of MCP into the AI agents built on its platform is anticipated to be revealed at the upcoming conference.
