Build Your Own AI-Powered DevOps Assistant with 7 AWS MCP Servers: Complete Guide + Source Code (2026 Edition)
Get the Complete Guide + Source Code: Click Here
What if you could manage and investigate your AWS infrastructure simply by asking Claude?
Imagine asking an AI assistant:
Show me every EC2 instance running in production.
Which IAM users don’t have MFA enabled?
Why did my AWS bill suddenly increase?
Who modified the Security Group before production went down?
Which EKS pods are stuck in CrashLoopBackOff?
Why is my production application slow?
Instead of manually navigating multiple AWS services, checking dashboards, executing CLI commands and searching through logs, imagine having an AI assistant that can retrieve information from different AWS services through specialized tools.
This is where Model Context Protocol (MCP) becomes interesting for DevOps and Cloud Engineering.
By combining MCP with Python, FastMCP and AWS boto3, we can build an assistant capable of investigating different parts of our cloud infrastructure.
In this article, I’ll walk through the architecture of an AI-powered DevOps Assistant built around seven independent AWS MCP servers.
1. Why AI-Powered DevOps?
Traditional DevOps workflows involve working with multiple tools and services.
For example, investigating a production issue might require:
Checking CloudWatch metrics.
Searching application logs.
Inspecting EC2 instances.
Checking EKS pods and deployments.
Reviewing CloudTrail events.
Investigating recent IAM or Security Group changes.
Each service provides useful information, but engineers often need to manually collect and correlate evidence.
An AI-powered DevOps assistant can simplify this investigation process.
Instead of remembering every AWS CLI command, you can ask a question in natural language.
The objective isn’t to replace DevOps engineers. It’s to reduce repetitive investigation and make infrastructure information easier to access.
2. Understanding the Architecture
The project follows this architecture:
Claude Desktop
↓
MCP Client
↓
MCP Server
↓
Tool Layer
↓
Service Layer
↓
Python + boto3
↓
AWS APIs
↓
AWS Infrastructure
Each component has a specific responsibility:
Claude Desktop: The AI client through which users ask questions.
MCP Client: Discovers available MCP servers, manages communication and invokes tools.
MCP Server: Exposes capabilities to the AI client.
Tool Layer: Defines the tools Claude can invoke.
Service Layer: Contains the AWS integration logic.
boto3: Python SDK used to communicate with AWS services.
This separation makes the architecture easier to understand, maintain, test and extend.
3. Why Seven AWS MCP Servers Instead of One?
Rather than creating one enormous MCP server responsible for everything, this project uses seven specialized servers.
Each server focuses on a specific AWS operational domain.
1. AWS Labs MCP: Infrastructure Inventory
This is the general-purpose infrastructure server.
It covers:
EC2 instances
VPCs and Subnets
Route Tables
Internet and NAT Gateways
Security Groups
S3 Buckets
Lambda Functions
ECR Repositories
EBS Volumes
Route 53
Example request:
“Show me every EC2 instance running in ap-south-1.”
Claude can invoke the relevant infrastructure tool instead of requiring you to construct an AWS CLI command manually.
2. CloudWatch MCP: Observability
CloudWatch is one of the first places engineers investigate when production applications experience problems.
This server covers:
Log Groups
Log Search
Metrics
CloudWatch Alarms
CPU Analysis
Idle Instances
Network Analysis
Example request:
“Show me the recent application errors and CPU utilization.”
Claude can retrieve relevant operational information through CloudWatch tools.
3. IAM MCP: Identity and Security
IAM security reviews often involve repetitive checks.
This server covers:
IAM Users
MFA Status
Roles and Policies
Administrator Accounts
Access Keys
Unused Roles
Permission Analysis
Example request:
“Which IAM users don’t have MFA enabled?”
This makes repetitive identity and access reviews easier to investigate.
4. CloudTrail MCP: Audit and Change History
When production suddenly stops working, identifying recent infrastructure changes is important.
CloudTrail MCP covers:
Recent AWS Events
User Activity
Security Group Changes
IAM Changes
Deleted Resources
Failed Login Events
Example request:
“Who modified the Security Group before the incident?”
The server can retrieve relevant CloudTrail events to support the investigation.
5. Cost Explorer MCP: Billing Intelligence
Cloud cost visibility is another important part of DevOps.
This server covers:
Monthly AWS Costs
Service-wise Spending
Cost Forecasting
Cost Spikes
Idle Resources
Budget Alerts
Example request:
“Why did my AWS spending increase this month?”
Claude can retrieve billing information and help identify potential cost optimization opportunities.
6. EKS MCP: Kubernetes Operations
Kubernetes troubleshooting frequently requires engineers to inspect several resources.
EKS MCP covers:
EKS Clusters
Pods
Deployments
Container Logs
Kubernetes Events
Node Health
Services and Ingress
Resource Usage
Example request:
“Show me the failing pods in my production namespace.”
This brings Kubernetes operational information into the MCP-powered workflow.
7. SSM MCP: Remote Operations
AWS Systems Manager provides capabilities for managing instances and operational tasks.
This server covers:
Managed Instances
Remote Diagnostics
Patch Compliance
Software Inventory
System Information
Session Manager
It uses controlled diagnostic operations rather than exposing arbitrary shell command execution.
4. How Multiple MCP Servers Work Together
This is where the architecture becomes particularly interesting.
Imagine asking Claude:
“Why is my production application slow?”
A possible investigation workflow could look like this:
Step 1: CloudWatch MCP
Retrieve application logs, metrics and alarms.
Step 2: EKS MCP
Inspect pod health, deployments and Kubernetes events.
Step 3: AWS Labs MCP
Investigate EC2 and infrastructure information.
Step 4: CloudTrail MCP
Retrieve recent infrastructure changes.
Step 5: Claude
Combine the available evidence and explain possible causes.
The important concept is that each server remains focused on its own domain while Claude can coordinate calls across multiple servers.
This is the difference between building an isolated MCP demonstration and exploring a multi-server AI-powered operational workflow.
5. The Technology Stack
The project uses several technologies:
Python: Main programming language used to develop the MCP servers.
FastMCP: Used to build and expose MCP tools.
boto3: Python SDK used to communicate with AWS services.
AWS CLI: Handles AWS authentication and configuration.
Claude Desktop: AI client that discovers and invokes MCP tools.
Docker: Used to explore containerization.
Docker Compose: Helps manage multiple server containers.
pytest: Used for automated testing.
The implementation starts with a simple Python MCP server.
You then learn how to expose functions using FastMCP, connect those functions to AWS through boto3 and gradually build the seven specialized servers.
6. From Local Development to Deployment
Building an MCP server locally is only the beginning.
The project also explores how to organize and prepare the seven servers for deployment.
Topics include:
Independent project structures.
Individual Dockerfiles.
Docker Compose.
Environment-based configuration.
Logging and audit trails.
Response caching.
Rate limiting.
Multi-account AWS profiles.
IAM security policies.
Troubleshooting.
Each server has its own project structure, making independent development, testing and updates easier to manage.
The repository includes deployment configurations, but real AWS deployment, EKS execution, Docker image builds and Claude Desktop integration have not been independently verified. These should be validated in a non-production environment before operational use.
7. Security: An Important Part of AI-Powered DevOps
Giving an AI assistant access to cloud infrastructure requires careful permission management.
You should never casually provide unrestricted AWS AdministratorAccess to an AI-connected application.
The project emphasizes:
Least-privilege IAM policies.
Separate permissions for individual servers.
Read-only operations by design.
Input validation.
Structured error handling.
Restricted SSM diagnostic commands.
Protection against arbitrary command execution.
Controlled AWS account and profile selection.
Security should be part of the architecture from the beginning, not something added after deployment.
8. What Does the Complete Project Include?
The project is accompanied by an actual source code repository and an 80-page implementation guide.
Here’s what you’ll receive:
Complete PDF Guide
80-page implementation walkthrough.
MCP fundamentals and architecture.
Step-by-step server development.
AWS integration and configuration.
Deployment and troubleshooting guidance.
Seven AWS MCP Servers
AWS Labs MCP.
CloudWatch MCP.
IAM MCP.
CloudTrail MCP.
Cost Explorer MCP.
EKS MCP.
SSM MCP.
Complete Source Code
Seven independent Python codebases.
88 implemented MCP tools.
FastMCP and boto3 integration.
Individual project structures.
Environment configuration examples.
Automated Testing
197 unit tests.
6 end-to-end tests.
Examples of mocked AWS interactions.
MCP testing resources.
Deployment and Security Resources
Dockerfiles.
Docker Compose configuration.
IAM security policies.
Architecture documentation.
Troubleshooting documentation.
The repository documents 197 unit tests and 6 end-to-end tests. These provide a foundation for understanding and exploring automated testing.
The important difference is that you receive both:
Knowledge: A structured walkthrough explaining the architecture and implementation.
Implementation: Actual Python source code for seven independent AWS MCP servers.
You can study the code, understand individual functions, modify existing tools and extend the project with additional integrations.
9. Who Should Explore This Project?
DevOps Engineers
Explore how AI can interact with infrastructure and simplify operational workflows.
AWS and Cloud Engineers
Understand how MCP connects AI assistants with AWS services.
Kubernetes Engineers
Explore EKS operations through specialized MCP tools.
SREs and Platform Engineers
Investigate AI-assisted monitoring, troubleshooting and infrastructure analysis.
Developers Learning MCP
Move beyond definitions and build practical MCP servers.
Engineers Exploring AI Automation
Understand how to connect AI assistants with external systems.
10. The Bigger Picture: AI + DevOps + MCP
DevOps engineering continues to evolve.
We’re already working with automation, Infrastructure as Code, GitOps, observability and cloud-native platforms.
MCP introduces another interesting possibility: connecting AI assistants to the tools and systems engineers already use.
Consider the possibilities:
AI-assisted infrastructure investigation.
Natural-language cloud queries.
Faster access to monitoring information.
Automated collection of operational evidence.
Cross-service troubleshooting workflows.
AI-assisted security and cost reviews.
These are practical areas worth exploring.
The real engineering challenge is not simply connecting an AI model to AWS. It’s designing the architecture, controlling permissions, validating tool outputs and making the system reliable.
That’s what makes this project interesting.
🚀 Build Your Own AI-Powered DevOps Assistant
I’ve created a complete learning and implementation package for engineers who want to explore this architecture hands-on.
What’s included?
80-page step-by-step implementation guide.
7 independent AWS MCP server codebases.
88 implemented MCP tools.
Complete Python source code.
FastMCP and boto3 integration.
197 unit tests and 6 end-to-end tests.
Docker and deployment configurations.
IAM security policies.
Architecture and troubleshooting documentation.
This is designed to take you from your first MCP server to exploring a complete multi-server AI-powered DevOps architecture.
🔗 Get the Complete Guide + Source Code:
https://topmate.io/arvindverma021/2242029
Don’t Just Learn MCP. Build Something With It.
Start with Python.
Build your first MCP tool.
Connect it to AWS.
Develop seven specialized servers.
Connect them to Claude Desktop.
Explore how AI can investigate your infrastructure.
Learn. Build. Explore. Extend.
About the Author
Arvind Verma | Senior DevOps Engineer
I share practical DevOps tutorials, AWS and Kubernetes implementations, infrastructure automation, production troubleshooting and engineering resources.
Follow me on LinkedIn:
https://www.linkedin.com/in/arvindverma021/
Tags
DevOps, AWS, Artificial Intelligence, Model Context Protocol, Kubernetes, Cloud Computing
#DevOps #AWS #AIDevOps #MCP #ModelContextProtocol #Kubernetes #EKS #Python #FastMCP #ClaudeAI #CloudEngineering #PlatformEngineering #SRE #DevSecOps #CloudAutomation #AIAgents #CloudNative #Docker #CloudMonitoring #GenerativeAI
Arvind Verma is free today. But if you enjoyed this post, you can tell Arvind Verma that their writing is valuable by pledging a future subscription. You won't be charged unless they enable payments.
© 2026 Arvind Verma
548 Market Street PMB 72296, San Francisco, CA 94104