Disclosure: This is a sponsored hands-on review. Prisma compensated the write-up and covered the Pro plan I used for testing. Opinions, timings, and “who it’s for” notes are from my own StackDuel test (9–12 Nov 2026). Some Prisma links are (referral) partner links — Vibe Coder's Life may earn a reward if you sign up. Product and docs links below are otherwise plain.
I wanted to know whether Prisma's newer platform can run a real TypeScript app — not a demo that works once on a laptop. So I built one.
StackDuel is a swipe game for people who build with AI: Claude Code or Cursor? Free tier or paid from day one? You pick blue or red, see the live split, and tool duels move a public ranking. It needs a real database, concurrent writes, a live feed and a host — exactly the things that break.
Over four days I put Prisma Postgres and Prisma Compute through deploy, preview branches, cold starts, backups and a 10,000-vote load run on a dedicated loadtest environment (so the public demo stayed clean). Votes stayed correct. The bill stayed tiny. Agent tooling was strong enough that I would put a real TypeScript app here — from a vibe-coded build through a production team app.
Live demo: StackDuel on Prisma Compute
TL;DR
- First deploy: 51 seconds. Git push to a live preview: about 2 minutes, with its own fresh database.
- Load: 10,000 votes in 51 seconds, 64 in parallel, p95 674 ms, zero server errors. Counters matched the vote table after every run.
- Bill at list prices: about 100,000 operations for the whole test — inside Free's 200,000 by usage. Starter is $10 + Compute usage if you want daily backups.
- Gaps: Prisma 8 is still a release candidate, no built-in end-user auth, point-in-time recovery is not available yet (on Prisma's roadmap), CLI rollback can be empty after redeploys.
- Fit: TypeScript apps that want app + Postgres + branch previews on one bill. Skip for now if you need auth/realtime/storage as one product, PITR today, or you can't accept an RC ORM.
What Prisma is (in this test)
Four pieces mattered:
-
Prisma ORM 8 — still a release candidate. I pinned
8.0.0-rc.33for the test;npmlatest moved to8.0.0-rc.34on 16 Nov 2026. Agents trained on Prisma 5–7 invent old APIs unless you give them skills and rules. - Prisma Postgres — managed PostgreSQL 17, billed by operations and storage, idle databases cost nothing.
- Prisma Compute — hosts TypeScript apps (Node / Bun / Next.js) next to the database in the same project.
-
Prisma Composer — describes app + database together so
prisma deployships both.
Why a vibe coder should care: one project, one bill, and a platform your AI agent can read and, with your approval, change. Coding agents can enroll under Console → Settings → Agents, work freely on preview branches, and ask before touching production.
What I built
StackDuel has 100 duels, 8 packs, 27 tools. No accounts — a cookie limits votes. Next.js 16 on Bun, one raw-SQL transaction per vote (Prisma 8 RC had no atomic increment I could use), rankings stream to open tabs.
The public URL is production for organic traffic. The 10,000-vote run hit a separate loadtest environment. If you open the live site and see on the order of a hundred votes, that is expected.
What held up
| Test | Result |
|---|---|
| First deploy | 51 s plan+apply (~2m 10s with Next build) |
| Git preview | ~2 min; fresh database per branch (migrations applied; production data not copied) |
| Load | 10k votes / 51 s / 64 parallel / p95 674 ms / 0 server errors |
| Cold start | ~1.0–1.3 s after ~18h idle |
| Live feed | 10 tabs / 10 min / 0 errors |
Takeaway: for this app's shape — concurrent writes, live rankings, preview branches — the boring parts held up. Votes were never lost or double counted.
What it costs (list prices, checked 12 Nov 2026)
Prisma bills operations plus storage. Paid plans include Compute requests only (5M / 20M / 100M). Memory, active CPU and outbound bandwidth are billed from the first unit, so Starter is $10 + Compute usage (same pattern on Pro / Business).
| Plan | Price | DB ops | Storage | Compute requests | Backups |
|---|---|---|---|---|---|
| Free | $0 | 200k | 2 GB | 1M | none |
| Starter | $10/mo + Compute usage | 1M | 10 GB | 5M | daily, 7 days |
| Pro | $49/mo + Compute usage | 10M | 50 GB | 20M | daily, 7 days |
| Business | $129/mo + Compute usage | 50M | 100 GB | 100M | daily, 30 days |
My four-day test used about 100,000 operations — inside Free by usage. One vote ≈ 5.6 billed operations. Cached rankings reads were ~0.014 ops; uncached ~1.0. Cache your hot reads.
Prisma covered Pro for the sponsorship; at list prices the Console showed about $0.02 of Compute memory on top of the plan fee for my run.
Spend limit: paid plans have a hard spend limit in the Console. It is off by default — turn it on and choose your cap before a paid deploy.
Things to know going in
- Prisma 8 is an RC. Pin exact versions. Install the Prisma 8 skills.
-
Prefer the hosted MCP at
https://mcp.prisma.io/mcp. On the RC I used, the localprisma mcppath did not start. - Previews ≠ data clones. Each branch gets its own fresh database with migrations applied; production data is not copied. Docs: branching.
- Point-in-time recovery is not available yet — on Prisma's roadmap. Paid plans have daily backups. Docs: backups.
- No built-in end-user auth. Pair Clerk / Better Auth / Auth.js if you need sign-in.
- CLI rollback can be empty after a few Composer redeploys. Prefer Git-connected deploys when you want older versions around.
- Destructive MCP actions waited for Console approval in my test — keep that on.
Five-step setup (Cursor, Claude Code, Codex)
- Account + spend limit — Try Prisma (referral). Free: 200k operations/month. On paid, turn the spend limit on (off by default).
bunx prisma auth login
- Scaffold Prisma 8 skills
npm create prisma@latest -- --skills claude,cursor,agents
# existing project:
npx prisma skills sync
# throwaway Postgres, no account (handy for agents):
npx create-db@latest
-
AGENTS.mdin the project root (Cursor and Codex read it). For Claude Code, addCLAUDE.mdcontaining@AGENTS.md. Pin exact versions; reject local MCP that runsnpx prisma mcp. -
Connect the agent — hosted MCP
https://mcp.prisma.io/mcp. Claude Code:npx plugins add prisma/prisma-pluginthen add the HTTP MCP. Cursor: marketplace plugin or MCP URL (not both). Codex:codex plugin marketplace add prisma/codex-plugin. Enroll under Console → Settings → Agents. -
Deploy —
bun run deploy(build +prisma deploy). Connect Git for previews; keep the default-branch workflow off production until you want every push to ship prod.
Who should pick it
Easy yes: TypeScript apps — vibe-coded through production team products — that want app + Postgres + branch previews on one bill, with strong agent tooling.
Skip for now if you need auth/realtime/storage as one product (Supabase), point-in-time recovery today, a non-TypeScript Compute backend, or you can't accept a release-candidate ORM.
Soft close
I would put a real TypeScript workload on Prisma Postgres + Compute, cache hot reads, turn on the spend limit, and watch operations for a week. If it fits, stay. The exit ramp is plain Postgres.
Longer write-up with the full comparison tables and the emailed AGENTS.md guide: Vibe Coder's Life — Prisma Postgres + Compute review.
Try Prisma: prisma.io (referral) · Pricing: prisma.io/pricing · Compute pricing notes: docs



Top comments (0)