Amazon Is Betting Big on MCP — and That Changes What the Standard Becomes

The Model Context Protocol started as an Anthropic internal standard and became an open protocol when the Linux Foundation took custody in late 2025. The interesting question was never whether other vendors would support it — it was which vendor would invest deeply enough to shape where the spec goes. That vendor is Amazon.

Context

Clare Liguori, AWS Senior Principal Software Engineer, is both a core MCP maintainer and the most visible enterprise voice at MCP’s evolution table. At the MCP Summit in New York City, she outlined AWS’s production deployment of MCP features — including features still in draft specification. AWS is effectively using its own infrastructure as the testing ground for emerging MCP concepts before they’re finalized, which gives Amazon direct influence over what becomes standard.

This is a calculated move. By the time other vendors implement a new MCP feature, AWS will have months of production data behind it. For enterprise cloud architects evaluating where to build their agentic AI stack, that matters: the feature set and behavior of MCP on AWS will lead.

The Detail

AWS has already shipped several concrete MCP extensions. The Tasks feature extends request timelines to support the longer operations that complex agentic workflows require — standard REST request/response patterns break down when an agent is doing multi-step reasoning across data sources. Elicitations add the ability for agents to request human clarification mid-task, a capability that makes agentic workflows safer in regulated or high-stakes environments.

Liguori indicated that AWS is actively developing webhooks, events, and notifications for MCP, positioning the protocol for reactive agent architectures — agents that respond to incoming events rather than only to user-initiated requests. This is a significant architectural expansion: always-on agents that integrate with existing enterprise event infrastructure rather than requiring humans to trigger them.

The expansion of Amazon Kiro AI, an MCP-powered development tool, beyond engineering teams to all job families signals Amazon’s read of where this goes. If MCP is only a developer tool, it’s infrastructure; if it reaches non-technical business users, it’s a platform shift. AWS appears to be positioning for the latter.

The broader context is noted by The New Stack: since Anthropic’s MCP was unveiled in late 2024, it has become the de facto method for connecting AI agents with tools and data. Enterprise interest has reached a fever pitch in 2026, and cloud providers are competing to be the best-supported platform for MCP deployment.

Source: The New Stack

Content created with AI assistance and reviewed for accuracy.