Adoption and constraints of agent-to-agent interaction standards

Determine how standardized agent-to-agent interaction frameworks and protocols—including Microsoft’s AutoGen, Anthropic’s Model Context Protocol (MCP), and Google’s Agent2Agent protocol—will be adopted and implemented in practice, and ascertain the extent to which their use will be constrained, given their complex interplay with and dependence on market forces.

Background

The paper argues that generative AI will reduce communication frictions by enabling assistant agents for consumers and service agents for businesses to communicate directly. Realizing this vision requires interoperable agentic interaction across entities.

While technical progress exists via emerging frameworks and protocols (e.g., Microsoft AutoGen, Anthropic MCP, Google Agent2Agent), the authors explicitly note uncertainty about how these advances will be adopted, implemented, or constrained due to market dynamics. This adoption question underpins whether markets evolve toward closed agentic walled gardens or an open web of agents, with downstream impacts on market power, advertising, payments, and product bundling.