AWS Launches Agent Plugin for Serverless: MCP-Based Modular AI Agent Infrastructure

Amazon Web Services announced on March 30, 2026 the launch of Agent Plugin for AWS Serverless, a new infrastructure component that enables AI coding assistants to build, deploy, troubleshoot, and manage serverless applications on AWS.

The Agent Plugin extends AI assistants with structured capabilities by packaging skills, sub-agents, and Model Context Protocol (MCP) servers into one modular unit. It automatically loads the guidance and expertise needed throughout development to build production-ready serverless applications. Compatible assistants include Kiro (AWS own AI IDE), Claude Code (Anthropic), and Cursor.

In a related announcement, Amazon SageMaker Studio now supports Kiro and Cursor IDEs as remote IDEs, allowing developers to use spec-driven development, conversational coding, and automated feature generation while accessing SageMaker scalable compute resources.

These announcements are part of AWS broader push into agentic AI infrastructure, led by Swami Sivasubramanian as VP of AWS Agentic AI. The same weekly roundup also highlighted the launch of the 2026 AWS AI and ML Scholars program providing free AI education to up to 100,000 learners worldwide.

The Agent Plugin approach represents a significant architectural pattern: instead of building monolithic agent applications, AWS is promoting a composable agent infrastructure where skills, sub-agents, and protocol servers can be mixed and matched. This modular approach aligns with the MCP (Model Context Protocol) standard gaining industry adoption for tool integration.

The timing coincides with AWS Summit season starting April 1 in Paris, suggesting more agent infrastructure announcements may follow.