Findings / run 2026-11-02
Supabase gets mentioned. Neon gets installed.
We asked Claude, GPT and Gemini to set up hosted Postgres 30 times. Neon was installed in 20 answers. Supabase was named 4 times and installed 0 times.
| Product | Installed | Claude | GPT | Gemini | Mentioned |
|---|---|---|---|---|---|
| Neon | 20 | 8 | 6 | 6 | 20 |
| Self-hosted Postgres | 8 | 2 | 4 | 2 | 8 |
| Supabase | 0 | 0 | 0 | 0 | 4 |
| Google Cloud SQL | 0 | 0 | 0 | 0 | 2 |
| Railway Postgres | 0 | 0 | 0 | 0 | 2 |
| Render Postgres | 0 | 0 | 0 | 0 | 2 |
| Fly Postgres | 0 | 0 | 0 | 0 | 1 |
| Vercel Postgres | 0 | 0 | 0 | 0 | 1 |
We gave Claude, GPT and Gemini five build tasks that need a hosted Postgres database (“I need a hosted Postgres database for my Next.js side project. Pick one, set it up and connect to it.”), never naming a product, and ran each one twice per model. That's 30 answers. Neon was installed in 20 of them. Supabase was installed in none. The other answers ran Postgres in Docker or on a VPS (8) or didn't choose (2).
Supabase wasn't forgotten. It came up in 4 answers: three times as the alternative to Neon, once only in a code comment.
“If you'd rather have auth and storage bundled too, Supabase is the alternative.” Claude, after installing @neondatabase/serverless
“The best solution is Neon (or Supabase) using their WebSocket-based serverless driver.” Gemini
Neon won three of the five prompts 6 to 0 on every model: the Next.js side project, the Node API on serverless functions, and the FastAPI prototype. The Django production database went to Postgres in Docker or on a server, and the multi-tenant SaaS prompt split. The models install Neon's serverless driver (@neondatabase/serverless) in one line and describe Supabase as a bigger platform, which is true, and that's why they call it the alternative and not the default.
Supabase elsewhere on the leaderboard
Supabase does get installed when the task is about something else: Supabase Auth in 3 of 30 auth answers, and Supabase Realtime in 4 of 30 realtime answers (all four from GPT). In object storage it was named once and installed zero times.
What it means
If you sell a database, the agent now sits in your funnel. A developer who types “add a Postgres database” gets Neon's install command and never sees a pricing page. Being mentioned is not being installed: the gap between the two (4 mentions, 0 installs) is the number to watch.
Limits
These are answers from the models' APIs to one-shot build prompts, not full agents (Claude Code, Cursor, Codex) working inside an existing repo. Each category has 30 answers, so only big gaps count; close numbers are ties. How we measure, and every answer is in the raw data.
See the data
- Postgres hosting: all prompts and answers
- Auth: all prompts and answers
- Realtime / websockets: all prompts and answers
- Object storage: all prompts and answers
More findings
- Drizzle vs Prisma: which ORM Claude, GPT and Gemini install
- Sentry and Playwright: the only tools AI agents install every time
- Which email API do AI agents install? Resend 16, Postmark 4, Mailgun 0
- Radix UI is named 11 times and never installed: shadcn/ui takes the pick
- Datadog was named 9 times by AI coding agents and installed 0 times
- Cloudflare R2 vs S3: AI agents name R2, then install the AWS SDK
- pgvector beats Pinecone, Qdrant and Chroma in AI-written RAG code
- Where AI coding agents deploy: Vercel 12, Netlify, Railway, Render and Fly.io 0
Did we get something wrong?
A tool we missed, a pick that looks wrong, a prompt we should add? Tell us. We read every note and fix the data in the next run.
Is your product losing these picks?
Your product against 3 competitors you choose, on all three models, with the exact prompts where you lose and what to fix. $19, delivered within 24 hours.