Microsoft Shares Its AI Agents Vision

Microsoft presented a vision for AI agents from different companies to work together through open standards. Company CTO Kevin Scott said the company’s focus is on driving industry-wide standards to enable collaboration between AI agents built by different vendors.

To this end, Scott revealed that Microsoft is supporting the Model Context Protocol (MCP), an open-source framework from Google-backed AI company Anthropic aimed at enabling communication and integration between AI models. According to the executive, MCP has the potential to enable AI agents to collaborate across products and platforms.

Scott pitched MCP as a key step toward building what he called the agentic web, comparing it to how early internet standards powered the growth of the internet. “It means that your imagination gets to drive what the agentic web becomes, not just a handful of companies that happen to see some of these problems first,” he noted.

The CTO also addressed limitations in current AI systems, noting most interactions today are short-lived and very transactional. However, meeting this challenge by boosting AI recall and memory retention from past interactions typically demands more computing power.

As a result, Scott revealed that Microsoft is working on enhancing memory in AI agents using an approach dubbed ‘structured retrieval augmentation’, which captures and organizes key snippets from past conversations rather than storing entire interactions. This process reduces the need for expensive full-memory processing.