v1.0.0 — 35 modules across 16 categories

Give the AI a repo worth reading

One command writes the docs, rules, prompts and checklists your AI assistant needs — for your stack, in every tool you use.

Node ≥18 · MIT · free forever · read the docs →

Pro · --idea

The one part that isn't free, and why

A hosted service reads your sentence against the full catalogue and returns a stack — then the CLI shows it for review, exactly like a preset. Nothing is written until you say yes. It costs money to run, so it costs money to use: $15/month, no trial, and nothing else in the tool is behind it.

No key of your own

The model call runs on our budget, not an OpenAI account you set up. That is exactly why it is the one paid flag — there is no free allowance to give away.

Reviewed before anything is written

The proposed stack lands in the same confirmation screen a --preset does. Decline it and no file is touched.

It can say no

A proposal naming a module the catalogue doesn't have, or two that conflict, is rejected rather than half-built. You fall back to the wizard, free as ever.

The rest stays free

Cancel and the CLI does not degrade. Every local command keeps working on the projects you already generated — only --idea stops.

$ npx ai-project-bootstrap --idea "a habit tracker with paid reminders"
 
◆ reading your idea against 35 modules…
 
Proposed stack
mobile Expo
backend Supabase
auth Supabase Auth
payments RevenueCat
analytics PostHog
 
Est. $25/mo · 2 usage-based
 

The proposal is always shown before any file is written.

How it works

Three questions in, a repo out

01
Answer

Which AI tools you use, then mobile or web, then whatever your stack actually needs. Categories with nothing installed are skipped.

or skip it — --preset web-saas
02
Review

Every preset shows what it filled before anything is written, and you can back out to a fully custom run. --dry-run prints the file list and stops.

est. cost is shown here, before you commit
03
Generate

Deterministic — no timestamps, no absolute paths, stable ordering. The same answers always produce byte-identical output.

re-run later; hand-edits are preserved
$ npx ai-project-bootstrap
◆ Which AI coding tools do you use?
■ Cursor ■ Claude Code □ GitHub Copilot □ Continue.dev □ Cline □ Roo Code
◆ What are you building?
● Mobile app ○ Web app ○ Both
◆ Backend
● Supabase ○ Firebase ○ FastAPI ○ NestJS ○ None
✔ 43 files written to ./my-app
Est. cost $51/mo (Supabase, Sentry) — 1 usage-based service not counted
Try it here

Build a stack, see the repo

Same catalogue the CLI reads. Pick anything and the file tree, the exact command and your third-party service bill update as you go.

Start from
Your stack
What are you building?pick one
Backendpick one
Authenticationpick one
Databasepick one
Cross-cutting featurespick any
Paymentspick one
Analyticspick any
Crash Reportingpick any
Notificationspick one
Storagepick one
Emailpick one
Monitoringpick one
CI/CDpick one
Testingpick any
Deploymentpick one
Generated filesbaseline only
docs/8 files
prompts/reusable
checklists/release.md
.github/workflows/ci.yml
.env.examplededuplicated
package.jsonmerged
AGENTS.mdCodex
GEMINI.mdGemini CLI
ai-project.config.jsonreplayable
What your stack will bill you
$0/mo
across the paid services you selected
ai-project-bootstrap is free and MIT-licensed. This is what Supabase, Sentry, Stripe and the rest charge you directly — the generator never takes a cut and never sees a bill.
Counted above — flat or freemium tiers
nothing paid selected yet
Billed on your usage — not counted
none selected
Free at typical usage
none selected
Nothing to bill — no vendor
none selected

An estimate, not a quote. Each figure was checked by hand against that vendor's own pricing page — hover a row for the date — and CI fails the build when any figure is more than six months old.

$ npx ai-project-bootstrap
the interactive wizard
Pro · coming nextincluded in the subscription

Keep talking after the first proposal

--idea is one question and one answer. Chat is the same service with the conversation left open — argue with the stack, ask what a module actually costs, and take the command when you're happy. On this site and inside Claude Code over MCP, on the same subscription as --idea.

One subscription

Chat is not a second bill. When it ships it is included in Pro, alongside --idea, at the same $15/month.

Two places, one service

The same conversation on this site and inside Claude Code over MCP, so you can ask from the editor you're already in.

It ends in a command

Every thread resolves to a selection you can run or a config you can commit. Talking is not the deliverable.

stack chat — signed in as [email protected]
Ask about the stack…
Pricing

Free is the whole CLI. Paid is one flag.

There is no free tier of Pro and no trial, because every Pro request spends real API budget. There is also no crippled free version — the line is simply whether a command talks to a server.

FreeMIT · forever
$0
No account. No key. Nothing phones home.
The wizard, 16 categories, 35 modules
All three presets and the habit-tracker archetype
add · upgrade · implement · review · analyze · doctor
Cost estimation, and rules for all eight AI tools
Works offline, and keeps working if you never pay
npx ai-project-bootstrap
Prosubscription only
$15/mo
or $149/year. No free tier, no trial.
Everything in Free, unchanged
--idea — describe the project, get a stack proposed
Runs on our API budget, no OpenAI key of your own
Chat and the Claude Code assistant when they ship
Subscribe →

Full comparison →

Start now

Nothing to install. Nothing to sign up for.

$ npx ai-project-bootstrapthe full wizard
$ npx ai-project-bootstrap --preset web-saas --yesa curated stack, non-interactively
$ npx ai-project-bootstrap --archetype habit-trackera running app, not just dependencies