The PropelKit AI Product Manager: Your Shortcut from Idea to Code

Run /pk:start, answer a few questions about your SaaS idea, and get a customized codebase ready to build on. Here's how PropelKit's AI PM works.

TA

Tanishq Agarwal

April 20, 2026 · 6 min read

Share:
The PropelKit AI Product Manager: Your Shortcut from Idea to Code

Most developers waste the first day of a new project on setup.

Install dependencies. Configure the database. Set up auth. Wire up Stripe. Set environment variables. Wonder why the webhook isn't firing. Repeat.

PropelKit's AI Product Manager skips all of that. Not by doing setup faster — by doing it right the first time, customized for your specific product.

Here's how it works.

What is the PropelKit AI PM?

The AI PM is a Claude Code agent built into PropelKit. You run it with a single command:

/pk:start

It then interviews you about your SaaS idea — in a real conversation, not a form — and uses your answers to customize PropelKit for your specific product.

The goal: get from "I have an idea" to "I have working code that matches my idea" in under an hour.

The Interview Process

The AI PM asks a sequence of questions designed to understand your product without wasting your time.

It starts broad:

"What are you building? Tell me the problem it solves and who you're building it for."

Then it gets specific based on your answers:

"You mentioned teams — do you need different permission levels within a team? For example, can some members invite others?"

"Will your users pay per seat, per usage, or a flat subscription?"

"Do you need file uploads? What types — documents, images, or both?"

It continues until it has a complete picture of:

  • Your product's core feature set
  • Your data model
  • Your payment structure
  • Your user types and permissions
  • Design preferences

Then it summarizes what it understood and asks you to confirm before generating anything.

What It Generates

Once confirmed, the AI PM generates or customizes:

Configuration files:

  • src/config/brand.ts — your product name, logo, contact email, social links
  • src/config/features.ts — feature flags for your specific needs (multi-tenancy, credits, GST, etc.)
  • src/config/theme.ts — color scheme and typography

Database:

  • Supabase migrations for your core tables
  • RLS policies for your specific permission model
  • Initial seed data

API scaffolding:

  • Stub API routes for your main resources
  • TypeScript types for your data model

UI starter:

  • Basic dashboard layout matching your product's mental model
  • Navigation structure based on your feature set

This isn't boilerplate output — it's actually customized for what you described. A project management SaaS gets different tables and navigation than an analytics SaaS.

Watch Claude your SaaS in a Week. Without the halucinations. Get PropelKit →

The Lovable Integration

One of the most powerful parts of the AI PM is the Lovable flow.

When you reach the design step, the AI PM asks whether you want to:

  1. Start with PropelKit's default UI
  2. Generate a custom UI with Lovable

If you choose Lovable, the AI PM generates a detailed prompt — already incorporating your brand colors, product name, and feature set — that you can paste into Lovable to get a custom design.

Lovable outputs React components that you drop into your PropelKit project. The AI PM has already set up the component structure so they slot in without conflicts.

Real Example: Building a Subscription Newsletter Platform

Let me walk through a real AI PM session.

Me: I want to build a newsletter platform for creators. Writers can publish posts, have subscribers, and charge for premium content with a paywall.

AI PM: Got it. A few clarifications:

  • Will writers host their own domain, or will everyone be on yourplatform.com/[username]?
  • Do you need team/collaboration features, or is this solo creator focused?
  • For the paywall — do you want Stripe subscriptions, one-time purchases, or both?

Me: Solo creators only. Custom subdomain optional. Stripe subscriptions for premium.

AI PM: And for discovery — will you have a public directory of creators, or is it purely invite/link-based?

Me: Public directory.

AI PM: Perfect. Here's what I'll set up:

  • Users table with creator profiles and subdomain routing
  • Posts table with premium/free flag
  • Stripe subscriptions for reader-to-creator billing
  • Subscriber table with access tracking
  • Public creator directory page
  • Creator dashboard with post editor, subscriber count, revenue

Is this right?

Me: Yes, also add an email notification when a new post is published.

AI PM: Got it, adding Resend email dispatch to the post creation flow. Generating now...

Three minutes later, I had running code with exactly that structure. No generic setup, no decisions I'd already made asked again.

Why This Matters for Vibe Coding

The AI PM is designed for how people actually build SaaS in 2026.

Most developers use Claude or GPT to write their code. But there's a massive gap between "clone a generic boilerplate" and "have a codebase that's already structured for your specific product."

Without the AI PM, your first dozen Claude conversations are spent figuring out your data model, setting up RLS policies, and getting your configuration right. With it, you skip all of that and start your first Claude conversation with "add the subscriber management UI to the dashboard."

The AI PM's context — your product description, data model, feature decisions — also stays in the .claude/ directory. Future Claude conversations pick up where it left off. Claude Code knows about your product without you re-explaining it every session.

How It Compares to Other Boilerplates

No other Next.js SaaS boilerplate has anything like this.

ShipFast, MakerKit, IndieKit — they all give you a starting point and documentation. Then you read the docs, make a bunch of decisions, and start customizing.

PropelKit's AI PM does the customization with you. It's the difference between a blank canvas and a canvas that's already painted with your product's color palette.

Getting Started

  1. Purchase PropelKit at propelkit.dev
  2. Clone the repo to your development machine
  3. Set up your environment (Supabase project, Stripe account, Resend API key — takes ~20 minutes)
  4. Run /pk:start in Claude Code
  5. Answer the questions — be specific about your product
  6. Review and confirm the generated spec
  7. Start building your actual feature

The 90% that's the same across every SaaS — auth, payments, email, multi-tenancy — is already done. The AI PM helps you understand and configure it. Then you focus on the 10% that's actually your product.

That's the whole idea.


PropelKit includes the AI PM, plus auth, Stripe, Supabase, and everything else you need to ship. Get PropelKit →

Ready to ship your SaaS?

PropelKit gives you auth, payments, AI tools, multi-tenancy, and more. Go from idea to revenue in a day.

Get PropelKit
Share:

Related articles