$0 to start6 services· 6 auto-onboard

$0 AI App

Technical solo devs building an AI-powered web app

Web SaaS foundation + Replicate for pay-per-inference model access

Same foundation as Web SaaS (hosting, DB, auth, errors) plus Replicate for state-of-the-art models without the GPU management headache. Start with pay-per-inference — Replicate bills per second of compute, so you pay nothing until you have users. Add Lemon Squeezy when you're ready to charge customers. Deliberately omits Resend and Cloudflare for simplicity; both can be added à la carte from the Web SaaS package.

What's in the box

Every service has a transparent affiliate disclosure. Most have none.

GitHub

✓ wizard

Source control, CI

Free tier: Unlimited private repos
No affiliate — SPARKi earns nothing here

Vercel

✓ wizard

Hosting, edge functions, streaming responses

Free tier: 100 GB bandwidth/mo
No affiliate — SPARKi earns nothing here
When to upgrade: Bandwidth > 100 GB/mo → Vercel Pro ($20/mo) ↗

Supabase

✓ wizard

User auth + database for model inputs/outputs

Free tier: 500 MB database, 50K MAU, 1 GB storage
No affiliate — SPARKi earns nothing here

Replicate

✓ wizard

Pay-per-inference GPU model hosting

Free tier: No free tier; pay per second of compute. $0 upfront — only pay when you have users.
💰💰 Replicate has an affiliate program — SPARKi earns ~25% if you sign up through a referral link (details unclear without direct page access)
When to upgrade: Heavy volume or want better rates → Replicate Deployments (dedicated instances) ↗

Sentry

✓ wizard

Error tracking for both web and model-call failures

Free tier: 5,000 errors/mo
No affiliate — SPARKi earns nothing here

Pico

✓ wizard

Uptime monitoring

Free tier: Free — SPARKi's sister project
No affiliate — Pico is SPARKi's own tool

How to set it up

6 of 6 services are onboarded automatically by SPARKi's wizard. The rest you set up by following the steps below.

  1. 1Create accounts on GitHub, Vercel, Supabase, Replicate, Sentry
  2. 2Copy tokens to .env: VERCEL_TOKEN, GITHUB_TOKEN, SUPABASE_ACCESS_TOKEN, REPLICATE_API_TOKEN, SENTRY_AUTH_TOKEN
  3. 3Add providers to sparky.config.yaml for your AI project
  4. 4Configure your Replicate rate: REPLICATE_COST_PER_PREDICTION_USD (if using a per-image priced model) or REPLICATE_RATE_PER_SECOND_USD (for compute-billed models)
  5. 5Restart SPARKi: pm2 restart sparky --update-env
  6. 6Verify: /sparki packages setup ai-app

Ready? SPARKi can onboard most of this for you.

Open SPARKi and run /sparki start ai-appfor <your-project> in a direct message. SPARKi will walk you through 6 services one at a time, verify each API key live, and write the config automatically.

Open SPARKiBrowse other packages