ReleaseBytes
aws AWS Big Data Blog ·

Federated Data Access Patterns for AI Agents with AWS Bedrock AgentCore

blogaiawsarchitectaws-bedrock
announcement

This post outlines reference architectures for federated data access patterns using AWS Bedrock AgentCore and the open Model Context Protocol (MCP). It demonstrates how AI agents can directly query data across disparate enterprise systems—including S3, Kinesis, and relational databases—via a uniform interface. This approach aims to break down data silos and overcome access gaps, enabling business users to gain insights through natural language queries without needing deep data engineering expertise. The article provides a solution overview, a detailed walkthrough, and design patterns, with source code available for deployment.

  • →Bridging Data Silos for AI Agents
  • →Federated Data Foundation with Bedrock AgentCore & MCP
  • →AI Agent Orchestration and Data Query Flow
  • →Design Patterns for Data Access Flexibility
Notes (4) ›
  • Bridging Data Silos for AI Agents

    Enterprise data is fragmented across various systems, each with distinct access patterns and expertise requirements. This creates data silos and an access gap, hindering business users and AI agents from accessing insights efficiently without significant data engineering investment.

  • Federated Data Foundation with Bedrock AgentCore & MCP

    The article proposes a federated data foundation utilizing the Model Context Protocol (MCP) and Amazon Bedrock AgentCore. This architecture enables AI agents to directly query data from diverse sources like S3, Kinesis, and relational databases through a uniform interface, democratizing access.

  • AI Agent Orchestration and Data Query Flow

    A detailed walkthrough describes how user natural language requests are authenticated and orchestrated by an Amazon Bedrock AgentCore agent. The agent routes queries via AgentCore Gateway to specialized MCP servers, which translate tool calls into actions against various batch, streaming, and relational data sources.

  • Design Patterns for Data Access Flexibility

    Three design patterns are presented to implement federated data access, balancing governance and flexibility. One pattern, "Catalog-first access," uses AWS Glue Data Catalog and Amazon Athena via an AWS Data Processing MCP server to allow agents to query S3-based sources effectively.

Read the original announcement →

https://aws.amazon.com/blogs/big-data/from-silos-to-insights-federated-data-access-patterns-for-ai-agents/

Related releases