Skip to main content

​
Overview

The aws-research command performs AWS-specific technology research using the AWS Knowledge MCP server to match project requirements to AWS services, architecture patterns, Well-Architected Framework guidance, Security Hub controls, and UK Government compliance.

​
Usage

arckit aws-research "<topic or use case, e.g. 'serverless data pipeline', 'EKS networking'>"

​
How It Works

This command delegates to the arckit-aws-research agent which runs as an autonomous subprocess. The agent makes 15-30+ MCP calls to gather authoritative AWS documentation:
  1. Reads project requirements - Extracts AWS service needs from functional/non-functional requirements
  2. Identifies AWS categories - Compute, data, integration, security, AI/ML, networking
  3. Uses MCP tools to research each category:
    • search_documentation - Find relevant AWS services and patterns
    • read_documentation - Extract detailed service information
    • get_regional_availability - Check eu-west-2 (London) availability
    • recommend - Get AWS-recommended architecture patterns
  4. Well-Architected assessment - Maps to 6 pillars (Operational Excellence, Security, Reliability, Performance, Cost, Sustainability)
  5. Security Hub mapping - Identifies security controls and compliance requirements
  6. UK Government compliance - G-Cloud, data residency, NCSC alignment
  7. Cost estimation - Monthly AWS costs with optimization strategies
  8. Generates architecture diagram - Mermaid diagram of AWS services
  9. Writes research document - Saves to projects/{project}/research/ARC-{PROJECT_ID}-AWRS-v1.0.md

​
AWS Knowledge MCP Server

The research agent uses the AWS Knowledge MCP server for authoritative AWS data:

​
MCP Tools Used

search_documentation
  • Searches AWS documentation for services matching requirements
  • Example: “serverless compute options for Node.js”
  • Returns: Lambda, Fargate, App Runner service docs
read_documentation
  • Fetches detailed AWS service documentation
  • Example: Read Lambda pricing, limits, features
  • Returns: Full service specifications
get_regional_availability
  • Checks service availability in specific regions
  • Example: Check if Amazon Bedrock available in eu-west-2
  • Returns: Service availability, regional constraints
recommend
  • Gets AWS architecture pattern recommendations
  • Example: “Three-tier web application pattern”
  • Returns: Reference architectures, best practices

​
Standalone Mode (Without MCP)

If MCP tools are unavailable, the agent falls back to:
  • WebSearch with site:docs.aws.amazon.com
  • WebFetch on AWS documentation URLs
  • Manual architecture pattern extraction

​
Output

The command generates:

​
AWS Research Document

projects/{project}/research/ARC-{PROJECT_ID}-AWRS-v1.0.md containing:
  • AWS services recommended per category (compute, data, integration, security, AI/ML)
  • Architecture pattern and reference from AWS Architecture Center
  • Well-Architected assessment - Alignment to 6 pillars
  • Security alignment - Security Hub controls, compliance
  • UK Government suitability - G-Cloud status, eu-west-2 availability, data classification
  • Estimated monthly cost - With Reserved Instances, Savings Plans, Spot optimization
  • Mermaid architecture diagram - Visual representation of AWS services
  • Next steps - Recommended follow-up commands

​
Summary

The agent returns a concise summary including:
  • AWS services by category
  • Architecture pattern name
  • Security and compliance status
  • UK Government suitability assessment
  • Estimated monthly cost range

​
AWS Service Categories

​
Compute

  • AWS Lambda - Serverless functions
  • AWS Fargate - Serverless containers
  • Amazon EC2 - Virtual machines
  • Amazon EKS - Kubernetes
  • AWS App Runner - Container apps

​
Data

  • Amazon RDS - Relational databases (PostgreSQL, MySQL, Aurora)
  • Amazon DynamoDB - NoSQL database
  • Amazon S3 - Object storage
  • Amazon Redshift - Data warehouse
  • Amazon Aurora - Cloud-native database

​
Integration

  • Amazon API Gateway - API management
  • Amazon EventBridge - Event bus
  • AWS Step Functions - Workflow orchestration
  • Amazon SQS - Message queue
  • Amazon SNS - Pub/sub notifications

​
Security

  • AWS IAM - Identity and access
  • AWS Secrets Manager - Secrets storage
  • AWS KMS - Encryption keys
  • AWS WAF - Web application firewall
  • Amazon Cognito - User authentication

​
AI/ML

  • Amazon Bedrock - Generative AI (LLMs)
  • Amazon SageMaker - ML platform
  • Amazon Textract - Document analysis
  • Amazon Comprehend - NLP
  • Amazon Rekognition - Image/video analysis

​
Well-Architected Framework Assessment

The research document maps AWS services to the 6 pillars:

​
1. Operational Excellence

  • Infrastructure as Code (CloudFormation, CDK)
  • Monitoring (CloudWatch, X-Ray)
  • CI/CD (CodePipeline, CodeBuild)

​
2. Security

  • Encryption at rest and in transit
  • IAM least privilege
  • Security Hub compliance
  • WAF and DDoS protection

​
3. Reliability

  • Multi-AZ deployment
  • Auto Scaling
  • Backup and disaster recovery
  • Health checks and failover

​
4. Performance Efficiency

  • Right-sizing instances
  • Caching (CloudFront, ElastiCache)
  • Database optimization
  • Serverless where appropriate

​
5. Cost Optimization

  • Reserved Instances and Savings Plans
  • Spot Instances for batch workloads
  • Graviton instances (ARM, 20% cheaper)
  • S3 Intelligent-Tiering

​
6. Sustainability

  • Graviton (lower carbon footprint)
  • Serverless (efficient compute)
  • eu-west-2 (renewable energy region)

​
UK Government Compliance

For UK Government projects, the research includes:

​
G-Cloud Framework

​
Data Residency

  • eu-west-2 (London) - UK data residency
  • Data sovereignty compliance
  • Brexit-aligned service agreements

​
NCSC Alignment

  • Cloud Security Principles compliance
  • Cyber Essentials Plus certified
  • IL3 (Impact Level 3) hosting available

​
Technology Code of Practice

  • Point 5 (Cloud First) - AWS as cloud provider
  • Point 6 (Make things secure) - Security Hub, encryption
  • Point 11 (Choose the right tools) - G-Cloud procurement

​
Cost Estimation Example

### Estimated Monthly AWS Cost

| Service | Specification | Monthly Cost | Optimization |
|---------|--------------|--------------|-------------|
| EC2 (t4g.medium) | 2 instances, eu-west-2 | £50 | Graviton, Reserved Instance (-30%) |
| RDS Aurora PostgreSQL | db.t4g.medium, Multi-AZ | £120 | Graviton, Reserved Instance (-40%) |
| Lambda | 10M requests, 512MB | £15 | Graviton runtime |
| S3 | 500GB storage | £10 | Intelligent-Tiering |
| API Gateway | 10M requests | £35 | - |
| CloudWatch | Standard monitoring | £8 | - |
| **Total (On-Demand)** | | **£238/month** | |
| **Total (Optimized)** | | **£165/month** | **-31% savings** |

​
Architecture Pattern Examples

​
Three-Tier Web Application

CloudFront → ALB → ECS Fargate → RDS Aurora
           ↓
         WAF

​
Serverless API

API Gateway → Lambda → DynamoDB
            ↓
         Cognito (Auth)

​
Data Pipeline

S3 → EventBridge → Step Functions → Lambda → Redshift
                                    ↓
                                  Glue

​
Integration with Other Commands

Input from:
  • requirements - Requires ARC-*-REQ-*.md
  • data - Uses data model for database selection
Output to:
  • diagram - Creates AWS-specific architecture diagrams
  • secure - Validates against Secure by Design principles
  • devops - Designs AWS CodePipeline CI/CD
  • finops - Creates AWS cost management strategy

​
Examples

​
Example 1: Serverless API Research

arckit aws-research "serverless REST API for public sector case management"
Researches:
  • API Gateway vs ALB + Lambda
  • DynamoDB vs Aurora Serverless
  • Cognito for authentication
  • Step Functions for workflows
  • Cost comparison vs EC2-based approach

​
Example 2: Data Platform Research

arckit aws-research "data analytics platform for NHS patient data"
Researches:
  • S3 data lake architecture
  • Glue for ETL
  • Athena vs Redshift for analytics
  • QuickSight for visualization
  • HIPAA-eligible services (BAA required)
  • IL3 hosting for NHS data (OFFICIAL-SENSITIVE)

​
Example 3: AI/ML Research

arckit aws-research "document classification using generative AI"
Researches:
  • Amazon Bedrock (Claude, Titan models)
  • SageMaker for custom models
  • Textract for document extraction
  • eu-west-2 availability (Bedrock limited)
  • AI Playbook compliance (if HIGH-RISK AI)

​
Resources

​
Next Steps

After completing AWS research:
  1. Create Architecture Diagram - arckit diagram with AWS services
  2. Security Review - arckit secure to validate against Secure by Design
  3. CI/CD Design - arckit devops for AWS CodePipeline
  4. Cost Management - arckit finops for AWS cost optimization strategy
  5. ADR - arckit adr to record AWS service selection decisions
  • research - Generic technology research with build vs buy
  • azure-research - Azure-specific research using MCP
  • diagram - AWS architecture diagrams
  • devops - AWS CodePipeline CI/CD
  • finops - AWS cost management
  • adr - Architecture decision records

Build docs developers (and LLMs) love

Get started for freeTalk to us