Microsoft wants AI agents from different companies to work together
What's the story
Microsoft's Chief Technology Officer, Kevin Scott, has revealed the company's vision of a future where AI agents from different companies can work together.
Speaking at Microsoft's headquarters in Redmond, Washington, ahead of the Build conference, Scott stressed that these AI systems are meant to do specific tasks on their own.
He also noted Microsoft's commitment to supporting industry-wide standards for agent collaboration from different manufacturers.
Protocol support
Open-source protocol for AI collaboration
Scott revealed that Microsoft is supporting the Model Context Protocol (MCP), an open-source effort from Google-backed Anthropic.
He likened MCP's ability to generate an "agentic web" to the effect of hypertext protocols on the internet in the 1990s.
Scott said, "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."
Memory enhancement
Aim to enhance AI agents' memory
Scott also touched upon the need for AI agents to have better memory.
He said that most of the current developments feel "very transactional." However, improving an AI agent's memory can be expensive due to the need for more computing power.
To address this, Microsoft is looking at a new approach called structured retrieval augmentation. It would see an agent pulling out short segments from every user interaction, thereby creating a roadmap of what was discussed.