Build a multi-account AI agent with AgentCore Gateway and MCP
Build a multi-account architecture that keeps each team's data in its own AWS account while giving AI agents a unified way to query across them. A central platform account runs the agent using Amazon Bedrock AgentCore Gateway and MCP, while line-of-business accounts expose their data as MCP servers with secure cross-account access and fine-grained authorization.
Enterprises strive to create AI agents capable of reasoning across data stored in multiple AWS accounts. Teams maintain data ownership by keeping each dataset in their respective AWS accounts, ensuring clear data management and deployment processes. However, granting an agent access to data from various accounts typically requires data replication or complex IAM configurations.
The objective is to maintain data localization while enabling agents to query across accounts seamlessly. This post outlines a multi-account architecture leveraging Amazon Bedrock AgentCore Gateway and Model Context Protocol (MCP) to achieve this goal. Amazon Bedrock AgentCore provides a scalable, secure platform for building, deploying, and operating agents.
A central platform account hosts the agent tier and LLM inference through Amazon Bedrock, while LOB teams expose their data and tools as MCP servers. The AgentCore Gateway in the platform account serves as a unified endpoint for agents to discover and invoke tools across all registered LOB accounts. The architecture comprises three layers: a central platform account, distributed LOB accounts, and the AgentCore Gateway as the integration layer.
The platform account houses the agent control plane, where agents connect to the platform's Gateway rather than individual LOB MCP servers. Amazon Bedrock runs LLM inference within the platform account, providing control over foundational models, guardrails, and cost tracking. Each LOB account hosts its MCP server, running on AgentCore Runtime, a serverless, framework-agnostic environment that maintains data isolation and exposes data and tools to the agent through a single, unified MCP endpoint.
Cross-account integration occurs through the Gateway and AgentCore Identity, enabling secure, fine-grained access control with Amazon Bedrock Guardrails and Policy in AgentCore.
Written by urgent.news from AWS Machine Learning's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
