AWS Startup Architecture and Cost Plan
Startup-sized AWS designs with service picks, monthly cost estimates at today and 10x, IaC skeleton, alarms, and CI/CD.
Before & after
Same prompt, same agent — the only difference is the skill.
Building for high concurrency on AWS with serverless is definitely possible. You'd likely use Lambda for compute and probably want a database like DynamoDB. API Gateway could handle the HTTP routing. You might need to think about caching and maybe a CDN for static content. Cost optimization is important for serverless, so you'll want to monitor how much you're spending. Consider using Infrastructure as Code with CloudFormation. There are a few different ways to structure this, so it depends on your specific requirements.
SERVERLESS ARCHITECTURE COMPONENT STACK → API Gateway (regional) - 10k concurrent = 10k concurrent connections - Burst throttle: 40k requests/sec (built-in) - Caching: 1hr TTL on read endpoints - WAF: rate limiting at 100 req/sec per IP → Lambda compute layer - Reserved concurrency: 6000 (covers 10k concurrent + spikes) - Memory allocation: 1024 MB (balance cost/speed) - Timeout: 30 sec (API calls only, no long jobs) - Ephemeral storage: 512 MB for temp files → DynamoDB tables (on-demand billing) - Primary: Users table (userId partition key) - Secondary: Sessions table (sessionId, TTL=24hr) - Auto-scaling: covers 10k concurrent reads naturally - Point-in-time recovery: enabled → Elasticache Redis cluster (6 nodes) - Session cache + user preferences - Failover: automatic → S3 + CloudFront - Static assets (SPA, images) - Cache policy: 1 year for versioned assets → CloudWatch Logs + X-Ray - Structured logging (JSON) - Distributed tracing on all Lambda calls CLOUDFORMATION SKELETON Included: parameter definitions, Lambda IAM roles (least privilege), VPC optional (Lambda-only simpler) COST ESTIMATE 10k concurrent, 60% peak utilization = $2800/month compute + $400/month data RISK MITIGATION Cold start: use provisioned concurrency on auth Lambdas only (most latency-sensitive) DynamoDB hot partitions: use sort key distribution if user activity skewed Billed duration: Lambda only charges for 100ms increments
About this skill
Your AWS bill is $300 a month for an app with forty users. Or someone on the team wants EKS and multi-region before you have revenue. Or you just need to know whether this job belongs in Lambda or Fargate. This skill designs AWS stacks that a 2-10 person team can actually run and afford. Every service is priced, and every choice comes with the threshold that would change it.
What it does
- Picks compute (Lambda, Fargate, ECS) and data stores (RDS Postgres, Aurora, DynamoDB) by workload shape and hard service limits, not by fashion
- Prices the design line by line at expected load and at 10x, and separates the fixed floor (NAT gateways, idle databases, logs) from usage cost
- Pushes back on premature complexity like EKS, multi-region, or DynamoDB for relational data, and explains what to use instead
- Writes IaC in your team's tool (CDK, Terraform, SAM, CloudFormation) with stateful resources protected from accidental deletion
- Sets up GitHub-to-AWS deploys with OIDC roles instead of long-lived keys, plus a short list of alarms worth paging on
- Reviews a pasted bill or template and ranks fixes by monthly dollars saved
What you get
You get an architecture you can defend. It includes a text diagram, a service-choice table with "switch when" triggers, a two-point cost table, an IaC stack layout with a skeleton of the riskiest resource, a CI/CD plan, alarms, a security baseline, and at most 3 questions that would change the design.
Best for
Founders, small engineering teams, and solo developers launching or right-sizing a product on AWS without a dedicated platform engineer.
SKILL.md preview
---
name: aws-startup-architecture-cost-plan
description: Designs AWS architectures sized for startups and small teams. It chooses compute, data, and networking by workload shape and service limits, prices the design line by line at current and 10x load, and delivers an IaC skeleton, CI/CD, alarms, and a security baseline. Use when the user says "what AWS setup should we use", "design our serverless backend", "Lambda or Fargate for this?", "write the CDK/CloudFormation/Terraform for this", "our AWS bill keeps going up", "why is the NAT gateway so expensive", "set up deploys to AWS from GitHub", or "we're moving off Heroku to AWS".
---
# AWS Startup Architecture and Cost Plan
… (sign up to view the full skill)More development skills
View all Development skills →Chaos Experiment and Game Day Planner
Plans chaos experiments and game days: hypothesis, blast radius, abort conditions, tool manifests, rollback, and post-run review.
Claude Code Cost and Latency Routing
Configure Claude Code Router with model tiers, explicit fallbacks, compatibility checks, and a cost-aware validation plan.
CI/CD Pipeline Builder with Gated Deploys
Generates CI/CD workflows from your repo's real stack: cached fast-fail checks, pinned actions, and staging-to-prod deploys with gates.
Terraform Modules and Safe Deployments
Get reusable Terraform modules, safe state migrations, explicit provider wiring, and CI gates tied to reviewed plans.