AWS Infrastructure MCP Server
Python · FastMCP · AWS · boto3 · Pydantic
A local FastMCP server that exposes 15 read-only AWS diagnostic tools to IDE AI agents, enabling natural-language infrastructure queries across CloudWatch, Lambda, DynamoDB, IAM and Bedrock.
Use case
Lets a developer ask their IDE copilot plain-English questions about AWS infrastructure — retrieve Lambda errors, inspect DynamoDB schemas, audit IAM policies for privilege escalation, or send logs to Bedrock for root-cause analysis — without leaving the editor or writing boto3 code.
Architecture
A single Python server built on FastMCP 3.x exposes tools over stdio so the IDE spawns it as a subprocess. Each tool wraps a focused, read-only boto3 call with Pydantic-validated inputs and a global credential scrubber that redacts access keys and secrets from every response. Authentication piggybacks on boto3's default credential chain. 57 unit tests mock the boto3 boundary so the suite runs without live AWS credentials.
View source on GitHub