Deployment & DevOps for MVPs

Purpose

This recipe deploys a full-stack MVP to production with CI/CD pipelines, environment variable management, error tracking, and uptime monitoring — covering five platforms (Vercel, Railway, Fly.io, AWS Amplify, Render) with platform-specific setup instructions, cost benchmarks, and monitoring integration. The output is a live application with automated deployments on every Git push, proper secret management, error alerting, and uptime monitoring. [src1]

Prerequisites

• Application codebase in Git — pushed to GitHub or GitLab with a working local development setup
• Environment variables documented — list of all API keys, database URLs, and secrets needed at runtime
• Deployment platform account — free account at Vercel, Railway, Fly.io, AWS Amplify, or Render
• Sentry account (recommended) — free at sentry.io for error tracking (5K errors/mo free)
• Node.js 18+ or your runtime installed locally for CLI-based deployments
• Custom domain (optional) — registered at Cloudflare, Namecheap, or Porkbun ($10-15/yr)

Constraints

• Vercel Hobby plan is non-commercial only. Any revenue-generating MVP requires Pro at $20/user/mo. [src1]
• Railway charges from first minute of usage — no free tier since 2024. Hobby: $5/mo base + $5 credit. [src2]
• Fly.io has no general free tier for new orgs. Budget minimum ~$2-5/mo for a basic app. [src3]
• AWS Amplify free tier: 5 GB stored, 1 GB served, 1000 build minutes/mo. New accounts get up to $200 in credits. [src4]
• Environment secrets must use platform-native secret management (not .env files committed to Git).
• Database connections from serverless functions require connection pooling to avoid exhausting connection limits.

Tool Selection Decision

Which path?
├── Frontend-only MVP (Next.js, Astro, React SPA)
│   ├── budget = free → PATH A: Vercel Hobby (non-commercial)
│   └── budget > $0 → PATH B: Vercel Pro ($20/mo)
├── Full-stack MVP (frontend + API + database)
│   ├── budget < $10/mo → PATH C: Railway Hobby ($5/mo + usage)
│   ├── budget $10-50/mo → PATH D: Railway Pro ($20/mo) or Render ($7-25/mo)
│   └── budget > $50/mo → PATH E: Fly.io or AWS Amplify
├── API/backend only (no frontend)
│   ├── budget < $10/mo → PATH C: Railway Hobby
│   └── budget > $10/mo → PATH D: Railway Pro or Fly.io
└── Already in AWS ecosystem
    └── PATH F: AWS Amplify (leverage existing credits/services)

Execution Flow

Step 1: Connect Repository and Initial Deploy

Duration: 10-15 minutes · Tool: Deployment platform dashboard or CLI

Connect your Git repository to the deployment platform. Each platform auto-detects frameworks and configures build settings.

Vercel (frontend or Next.js full-stack):

# Dashboard: New Project → Import Git Repository → select repo
# CLI alternative:
npm i -g vercel
cd your-mvp-project
vercel --prod

Railway (full-stack with database):

# Dashboard: New Project → Deploy from GitHub repo
# CLI alternative:
npm i -g @railway/cli
railway login
railway init
railway up

# Add database:
railway add --database postgres
# Connection string auto-injected as DATABASE_URL

Fly.io (global edge deployment):

curl -L https://fly.io/install.sh | sh
fly auth login
fly launch    # Follow prompts
fly deploy

# Add PostgreSQL:
fly postgres create --name your-app-db
fly postgres attach your-app-db

Verify: Application loads at the platform-provided subdomain (e.g., your-app.vercel.app, your-app.up.railway.app, your-app.fly.dev). [src1] [src2] · If failed: Check build logs. Common causes: missing build command, wrong output directory, Node.js version mismatch.

Step 2: Configure Environment Variables

Duration: 5-10 minutes · Tool: Platform dashboard or CLI

Store all secrets as platform environment variables. Never commit secrets to Git.

# Vercel
vercel env add DATABASE_URL production
vercel env pull .env.local  # For local dev

# Railway
railway variables set DATABASE_URL="postgresql://..."

# Fly.io (encrypted, values hidden after setting)
fly secrets set DATABASE_URL="postgresql://..."

# AWS Amplify: App settings → Environment variables

Verify: Application starts with all env vars loaded. Check logs for "undefined" errors. · If failed: Variable names are case-sensitive. Redeploy after adding variables.

Step 3: Set Up CI/CD Pipeline

Duration: 10-15 minutes · Tool: Platform settings + GitHub Actions

All platforms include auto-deploy on push. Add testing and quality gates with GitHub Actions.

# .github/workflows/ci.yml
name: CI
on: [push, pull_request]
jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: 20
          cache: 'npm'
      - run: npm ci
      - run: npm run lint
      - run: npm test

Vercel creates preview deployments automatically for every PR. Railway and Render support staging environments via branch-based deployments.

Verify: Push a commit — CI tests run, preview deployment created, merge to main triggers production deploy. [src1] [src5] · If failed: Check GitHub Actions logs. Verify platform webhook is connected.

Step 4: Add Monitoring and Error Tracking

Duration: 10-15 minutes · Tool: Sentry + BetterStack/UptimeRobot

# Install Sentry SDK
npm install @sentry/node @sentry/profiling-node

# For Next.js (wizard auto-configures):
npx @sentry/wizard@latest -i nextjs

Add a health check endpoint at /api/health that verifies database connectivity and returns JSON status.

Set up uptime monitoring: BetterStack (free: 10 monitors, 3-min intervals) or UptimeRobot (free: 50 monitors, 5-min intervals).

Verify: Trigger a test error in Sentry — appears in dashboard within 30 seconds. Uptime monitor shows "UP". · If failed: Check Sentry DSN env var. Verify monitor URL is externally accessible.

Step 5: Configure Custom Domain and SSL

Duration: 5-15 minutes · Tool: Platform dashboard + DNS registrar

# Vercel: vercel domains add yourdomain.com
# Railway: Dashboard → Settings → Domains → Add
# Fly.io: fly certs create yourdomain.com
# Render: Dashboard → Settings → Custom Domains

# DNS records at registrar:
# A     @     [platform IP]
# CNAME www   [platform target]

Verify: https://yourdomain.com loads with valid SSL. HTTP redirects to HTTPS. · If failed: DNS propagation takes up to 48 hours. Check with dnschecker.org.

Step 6: Set Up Log Aggregation (Optional)

Duration: 10 minutes · Tool: Platform-native logs or external service

All platforms provide built-in log viewers. For advanced needs, configure log drains to Datadog, Axiom, or BetterStack.

# Vercel: vercel integrations add datadog
# Railway: Dashboard → Service → Logs (built-in)
# Fly.io: fly logs (or fly logs --json for parsing)

Verify: Generate requests and confirm log entries appear within 60 seconds. · If failed: Ensure app outputs to stdout/stderr, not file-based logging.

Output Schema

{
  "output_type": "deployed_mvp_infrastructure",
  "format": "live URL + CI/CD pipeline + monitoring",
  "columns": [
    {"name": "production_url", "type": "string", "description": "Custom domain or platform subdomain", "required": true},
    {"name": "platform", "type": "string", "description": "Hosting platform used", "required": true},
    {"name": "cicd_status", "type": "string", "description": "CI/CD pipeline status", "required": true},
    {"name": "preview_url_pattern", "type": "string", "description": "URL pattern for preview deployments", "required": false},
    {"name": "sentry_dsn", "type": "string", "description": "Sentry DSN for error tracking", "required": false},
    {"name": "uptime_monitor_url", "type": "string", "description": "Uptime monitoring dashboard URL", "required": false},
    {"name": "monthly_cost", "type": "number", "description": "Monthly infrastructure cost in USD", "required": true},
    {"name": "ssl_status", "type": "string", "description": "SSL certificate status", "required": true}
  ],
  "expected_row_count": "1",
  "sort_order": "N/A",
  "deduplication_key": "production_url"
}

Quality Benchmarks

If below minimum: Focus on CI/CD pipeline first, then add monitoring. A working deploy pipeline with tests is more valuable than monitoring a broken process.

Error Handling

Cost Breakdown

Anti-Patterns

Wrong: Hardcoding secrets in source code

Committing API keys, database URLs, or secrets directly in source code. Even private repos can leak — credentials in Git history persist forever even after file deletion. [src1]

Correct: Use platform-native secret management

Store all secrets as platform environment variables. Use .env.local for development only with .gitignore protection. Audit with git log --all -S "sk_live".

Wrong: No health checks or monitoring on production

Deploying an MVP and only discovering downtime when users complain. Without monitoring, silent failures go undetected for hours. [src6]

Correct: Health check endpoint + uptime monitor from day one

Add /api/health that checks database connectivity. Configure UptimeRobot or BetterStack (both free) to ping every 5 minutes with Slack/email alerts.

Wrong: Using same database for dev and production

A misconfigured migration or accidental deletion in development destroys production data.

Correct: Separate databases per environment

Use separate database instances for dev, staging, and production. Railway and Render auto-create per-environment databases.

When This Matters

Use this recipe when the agent needs to take a working MVP codebase from local development to a production deployment with CI/CD, environment management, and monitoring. Requires the application code to exist and work locally. This recipe handles infrastructure, not application development.