Choose Backend

Choose Backend

The decision tree. No "best" answer — depends on what your app does.

Decision matrix

Detailed pros/cons

Supabase ⭐ default for indie apps

Pros:

• Postgres (real database, SQL access)
• Built-in auth (email, magic link, OAuth, Apple Sign In)
• Realtime via Postgres LISTEN/NOTIFY
• Storage (S3-compatible)
• Edge Functions (Deno-based)
• Self-hostable (you own your data)
• Generous free tier ($0 up to 500MB DB, 1GB storage)

Cons:

• Realtime is OK but not Convex-tier
• Scaling beyond ~$200/mo means custom Postgres
• Edge Functions are Deno (not Node) — some libraries don't work

Best for: 80% of iOS apps. Default unless you have specific reason otherwise.

Convex

Pros:

• Realtime is the BEST in class (Cursor uses it)
• Functions = code (no SQL needed)
• Built-in optimistic updates
• Schema-as-code (no migrations)

Cons:

• Vendor lock-in (data isn't trivially portable)
• Smaller ecosystem
• Pricing more aggressive at scale ($25/mo + usage)

Best for: Realtime collab apps, multiplayer games, agentic UIs.

Firebase (Firestore)

Pros:

• Mature, reliable, Google
• Excellent client SDKs (offline-first built in)
• Free tier is generous
• Auth + Storage + Functions all integrated

Cons:

• NoSQL (Firestore) — joins are painful, queries limited
• Vendor lock-in (BigQuery export is one-way)
• Functions are slow cold-start
• Pricing can spike unexpectedly with bad query patterns

Best for: When you want Google-grade reliability and don't need SQL.

Cloudflare Workers + D1 + R2

Pros:

• Cheapest at scale (free tier covers most apps)
• Edge-deployed for low-latency global workloads
• Hyperdrive for connecting to your own Postgres
• D1 (SQLite) for small data, R2 for files

Cons:

• More plumbing (auth needs Lucia / Better Auth)
• D1 has scale limits (~10GB)
• Workers have CPU time limits per request

Best for: AI apps with heavy server logic, image processing, edge functions.

Just BaaS-free (server-side from RN direct to APIs)

For some apps you don't need a backend at all:

• App is offline-first (drizzle-orm + expo-sqlite)
• Auth via Apple Sign In (no email/password storage)
• Subscriptions via RevenueCat (handles backend for you)
• Sync via iCloud (CloudKit via expo-cloudkit)

This is the MINIMUM viable backend: zero servers. Works for journaling, habit tracking, calculators, single-player games.

Decision flow

Copy code
Is the app multi-user?
├─ No → No backend needed. Use SQLite + iCloud sync.
└─ Yes → Continue

Does it need realtime collab?
├─ Yes → Convex
└─ No → Continue

Will you ever migrate or sell?
├─ Need data portability → Supabase
└─ Don't care → Firebase or Supabase

How much budget?
├─ $0/mo target → Cloudflare Workers + D1
├─ <$100/mo → Supabase or Firebase free tier
└─ Scale-ready → Supabase Pro or custom

A sensible default

Start with the smallest option that fits the product and validate current pricing, regions, limits, data-processing terms, and exit path before creating an account. For many relational Expo apps, Supabase is a reasonable option because it offers Postgres, Auth, and an edge-function boundary for keeping private AI-provider keys off device.

Pair with

• add-supabase-auth (next skill) for auth setup
• add-trpc if going custom backend
• subscription-server-validation for RC + your backend integration