Build a multi-account AI agent with AgentCore Gateway and MCP
AWS shows how to build an AI agent that can query data across many accounts without moving it, using AgentCore Gateway and MCP.
Why Now
AWS released a guide on September 24 2026 for building multi‑account AI agents with AgentCore Gateway and Model Context Protocol (MCP).
What Happened
The architecture uses a central platform account that hosts the agent and LLM inference via Amazon Bedrock, while line‑of‑business (LOB) accounts expose data as MCP servers. AgentCore Gateway provides a single endpoint for tool discovery, authentication, and fine‑grained authorization across LOBs. The setup keeps data in its owning account and only streams needed data at query time.
Why It Matters
This lets enterprises keep strict data ownership and isolation while still enabling AI agents to reason over distributed data, reducing data replication and simplifying governance. It also centralizes cost tracking and model management in the platform account.
The Limitation
The guide focuses on a single agent scenario; scaling to many agents or very large data volumes may require additional configuration and cost considerations.
What You Can Do
Try setting up a small multi‑account test with AgentCore Gateway and MCP to see how data stays in place while an agent can access it.
Source
Read original sourceWhy we picked this
Describes multi-account AI agent architecture using Bedrock AgentCore, core AI integration.