AI agents should act proactively, make plans, leverage context, take real-world actions, and create their own feedback loops to be effective.

We're collaborating with partners like Anthropic to address complex enterprise challenges on top of the MCP protocol, such as agent identification and building a robust entitlements infrastructure.

MCP has rapidly gained traction because it fills a crucial niche in the open agentic web ecosystem, functioning similarly to HTTP by allowing sophisticated actions without imposing strict payload opinions.

We're committing to making MCP the standard protocol for agentic communication between agents and services, enabling seamless integration of Microsoft's first-party platform components.

Agents must adhere to the same data protection and rights as any other data, ensuring security and preventing credential theft.

The agentic web is emerging as a collaborative ecosystem where agents can perform increasingly complex tasks, driven by open protocols and services.

For agents to function effectively, they require memory and tools, which necessitates proper entitlements and data governance to ensure security.

Agents will likely see faster adoption in enterprises than in consumer markets, as CIOs demand interoperability between systems.

AI agents are often defined as systems that perform complex planning and interact with external systems, but this definition overlaps with many current language models.