All work
AICloudPlatformDesign + implementation · Celebal Technologies · 2025

AWS Observability MCP Server

A Model Context Protocol server that gives AI agents safe access to AWS CloudWatch, secured with Cognito OAuth 2.0 and JWT validation.

Problem

AI agents needed access to AWS observability data. Exposing CloudWatch to an LLM without identity and scoping is a security risk.

Solution

An MCP server built with FastMCP that puts CloudWatch tooling behind Amazon Cognito OAuth 2.0 and JWT validation, so agents get scoped, auditable access.

Outcome

A reusable observability surface for agents that is authenticated, scoped, and safe to put in front of an autonomous system.

OAuth 2.0
Auth
MCP
Protocol
JWT-verified
Tokens
Tech stack
FastMCPAmazon CognitoJWTCloudWatchPython

Why MCP

The Model Context Protocol gives agents a standard way to call tools. Wrapping CloudWatch as an MCP server makes observability a swappable capability instead of custom glue.

Security model

Cognito issues OAuth 2.0 tokens, and the server validates the JWT on every call before it runs a CloudWatch query. Access is tied to an identity and logged.

Related blueprint
MCP Observability Gateway