AWS outlines a metadata-driven data-lake MCP server on Amazon ECS Fargate
Published November 6, 2026, AWS’s design uses a serverless Model Context Protocol (MCP) server on Amazon ECS with AWS Fargate to give AI agents access to an enterprise data lake. Dataset onboarding uses a metadata JSON file, while the tools cover SQL queries, semantic search, and policy search.

Full Report
AWS describes a serverless MCP server for AI-agent access to an enterprise data lake, built on Amazon ECS and AWS Fargate. The architecture has four layers—client, MCP server, tools, and knowledge-base refresh—and any MCP-compatible client can discover and invoke tools through a standardized `/mcp` endpoint.
To add a dataset, the design uses a metadata JSON file rather than per-dataset integration code; new metadata is discovered on the next pipeline sync. Available tools include SQL querying through Amazon Athena and Amazon Redshift, semantic search through Amazon Bedrock Knowledge Bases, and policy search.
The security design applies Amazon Bedrock Guardrails to every model interaction and limits Athena tools to SELECT queries, rejecting write operations. The MCP server sits in private subnets behind an internal Application Load Balancer and is not internet-reachable. The client layer uses Amazon Cognito for JWT authentication and Amazon DynamoDB for session storage.
The authors report end-to-end interaction latency of 15–20 seconds in testing. This guide focuses on data-lake access; a separate AWS MCP guide published April 14, 2026, described a three-service application with a Gradio-based UI, an Amazon Bedrock-powered AI agent, and a FastMCP server, also running on Amazon ECS with AWS Fargate.
Why This Matters
The design makes dataset onboarding metadata-driven and exposes query and search tools through one MCP endpoint, giving readers concrete integration details to assess for data-lake agents. It also specifies network isolation, authentication, query restrictions, and the authors’ reported test latency.
Evidence
Start with the primary evidence, then review supporting sources and context.
Primary Evidence
1Serverless MCP server for AI agent data lake access on Amazon ECS Fargate
Aishwarya Pawar
“Learn how to build a serverless Model Context Protocol (MCP) server on Amazon ECS with AWS Fargate that gives AI agents secure, metadata-driven access to an enterprise data lake. Exposing a new dataset requires only a JSON file, with no per-dataset integration code.”
Recommended signals
Recommended by brand, topic, and recent related coverage