Agent leaderboards / All sectors / Databases

Databases: which databases coding agents choose

Neon takes about 46% of database picks, Supabase about 37%.

732 runs24 apps3 agents5 personasupdated 2026-10-14

The interactive board, open on databases. Open it full page · this page as Markdown

Read this leaderboard as textrankings, key learnings, method

Key learnings

We asked three agents to pick a database for 24 codebases, 732 runs in all. We asked in different words, and as five different kinds of people. Behind the top two, Turso took about 7%, and everything else came in lower.

95 vs 68

The agent changed which one came first

Claude Code and Cursor picked Neon most often. Codex went the other way, choosing Supabase 95 times against 68 for Neon.

280 of 359

The kind of app decided the pick

On apps with their own server, Neon won 280 of 359 runs. On apps with no server, Supabase won 111 of 142. On new apps built from a starter, Supabase won 144 of 231.

39 of 72

Wording moved more than half the cases

In 39 of 72 cases the runs didn't all land on the same product. A case here is one codebase with one agent, asked several times over in different words and as different people.

0 of 138

One host came up often and never won

Railway came up in 138 plans and was never picked.

  • The one enterprise codebase already ran on Azure, and Azure Database for PostgreSQL Flexible Server won 21 of its 24 runs.
  • Asked as senior engineers, the agents picked Neon in 98 of 113 runs.
  • SQLite came up in 525 plans and won none, but it is an embedded engine with nothing to sign up for, so that isn't a loss.
  • Across all 732 runs, the agents wrote it themselves just once.
Explore every run in the interactive board

The ranking732 runs

ProductWinsShare
1 Neonneon.tech 334 46%
2 Supabasesupabase.com 269 37%
3 Tursoturso.tech 48 7%
4 Azure Database for PostgreSQL Flexible Serverazure.microsoft.com 21 3%
5 Firebasefirebase.google.com 17 2%
6 Google Sheetsgoogle.com 10 1%
7 Netlify Formsnetlify.com 6 1%
8 Formspreeformspree.io 4 1%
9 Render Postgresrender.com 4 1%
10 Azure SQL Databaseazure.microsoft.com 3 0%
11 Sanitysanity.io 2 0%
12 PostgreSQLpostgresql.org 2 0%
13 Cloudflare KVcloudflare.com 2 0%
14 Amazon S3aws.amazon.com 2 0%
15 Cloudflare Durable Objectscloudflare.com 1 0%
16 Cloudflare D1cloudflare.com 1 0%
17 SQLite + Supabase 1 0%
18 Aivenaiven.io 1 0%
19 Amazon RDS for PostgreSQLaws.amazon.com 1 0%
20 Built in-houseoutcome 1 0%

By agent, by persona, by wording

By agent

Claude Code380 runsNeon · 184then Supabase · 120
Codex · GPT-6.1 Sol181 runsSupabase · 95then Neon · 68
Cursor · Grok 4.7171 runsNeon · 82then Supabase · 54

By persona

Vibe coder417 runsSupabase · 193then Neon · 150
Junior developer163 runsNeon · 74then Supabase · 71
Senior engineer113 runsNeon · 98then Turso · 6
Enterprise team24 runsAzure Database for PostgreSQL Flexible Server · 21then Azure SQL Database · 3
General15 runsNeon · 12then Supabase · 2

By what the ask stressed

The plain ask725 runsNeon · 331then Supabase · 267

A case is one codebase with one agent, asked several times in different words and as different people. 39 of 72 cases did not hold to a single database.

How this was measured

Every number on this page comes from a controlled experiment. We took 24 small applications, asked 3 coding agents (Claude Code, Codex (GPT-6.1 Sol), Cursor (Grok 4.7)) to pick a database for each of them, in several wordings and as a vibe coder and junior developer and senior engineer and enterprise team and general, and let the agent choose the product. Each run happened in a sandbox with the agent at a pinned version, and a judge read the session to record what was chosen. That is 732 runs. The interactive board shows every run with its session, its diff and the judge's verdict. A simulated user stood in for the owner of the codebase: it read the agent's plan and had to approve it before any code was written; it sent the agent back at least once in 361 runs. Read the methodology and the publications.

If you sell in this sector: what these numbers mean for a vendor.

Open the interactive boardThis page as Markdown

Other sectors