PM·AI — Cursor for Product Managers

Turn raw user feedback into prioritized feature specs in seconds.

→ Live Demo

---

What is this?

Inspired by Y Combinator's Spring 2026 Request for Startups: "Cursor for Product Managers".

Cursor and Claude Code are great at helping engineers build software once it's clear what needs to be built. But figuring out what to build is the hard part — and nobody has built AI tooling for that yet.

PM·AI solves that. Paste in raw user feedback (interview notes, app reviews, support tickets, survey responses) and the app:

1. Clusters feedback into key themes using AI
2. Scores each theme by frequency × severity to give a priority ranking
3. Generates a full product spec — success metrics, user stories, UI changes, data model changes, and atomic dev tasks ready to paste into Cursor or Claude Code

---

Features

• 📋 AI Theme Analysis — clusters raw feedback into 4–6 prioritized themes
• 🎯 Product Context — add your product description for more specific, relevant specs
• 📊 Priority Scoring — frequency × severity matrix with Critical / High / Medium labels
• 🔀 Drag to Reorder — override AI priority with your own judgment by dragging cards
• 📄 Spec Generator — one-click full PRD with dev task breakdown
• 💾 Print / Save PDF — export any spec as a clean PDF
• 📋 Copy as Markdown — paste directly into Cursor, Claude Code, Notion, or Linear
• 🕓 History — last 10 analyses saved locally in your browser
• 🎭 Live Demo — pre-loaded Notion feedback so visitors see real output instantly

---

Tech Stack

Layer	Technology
Frontend	Next.js 15 (App Router) + TypeScript
Styling	Inline CSS + custom animations
Fonts	Playfair Display + JetBrains Mono
AI	OpenRouter (free tier)
Drag & Drop	@dnd-kit/core
Analytics	Vercel Analytics
Deployment	Vercel (free tier)
Database	None — stateless, localStorage for history

---

Why OpenRouter?

• Access to dozens of free LLMs with a single API key
• OpenAI-compatible API — swap models with one line change
• No credit card required to get started

Recommended free models (change one line in route.ts):

Model	Notes
meta-llama/llama-3.3-8b-instruct:free	Best for structured JSON output
google/gemini-2.0-flash-thinking-exp:free	Strongest reasoning
mistralai/mistral-7b-instruct:free	Fast and lightweight

---

Running Locally

Prerequisites

• Node.js 18+
• A free OpenRouter API key at openrouter.ai

Setup

# Clone the repo
git clone https://github.com/arkham2424/pm-ai.git
cd pm-ai

# Install dependencies
npm install

⚠️ Required: Create your environment file

This app needs a free OpenRouter API key to run. Without it, all analysis requests will fail.

1. Go to openrouter.ai → Sign up → Keys → Create Key
2. Create a file called .env.local in the root of the project (same level as package.json)
3. Add this line inside it:

OPENROUTER_API_KEY=sk-or-your-key-here

.env.local is already in .gitignore — it will never be committed. Every developer cloning this repo needs to create their own.

# Start the dev server
npm run dev

Open http://localhost:3000 — click View Demo to see it in action instantly without needing any feedback data.

---

Project Structure

pm-ai/
├── app/
│   ├── page.tsx              # Main UI — upload, analysis, spec view
│   ├── layout.tsx            # Root layout + OG meta tags + Analytics
│   └── api/
│       └── analyze/
│           └── route.ts      # AI API route (keeps key server-side)
├── public/
│   └── og.png                # OG image for link previews
├── .env.local                # Your API key (never committed)
└── package.json

---

How the AI Pipeline Works

User Feedback (raw text) + Product Context (optional)
        ↓
AI — Theme Extraction
  → Clusters feedback into 4-6 themes
  → Scores each by frequency (1-10) and severity (1-10)
  → Returns priority_score = frequency × severity / 10
        ↓
Ranked Theme List (draggable — reorder by your own judgment)
        ↓
AI — Spec Generation (on click)
  → Problem statement + why now
  → Success metrics
  → User stories
  → UI changes
  → Data model changes
  → Dev tasks with time estimates
        ↓
Export: Markdown (Cursor/Claude Code) or PDF

---

Inspiration

Built in response to Y Combinator's Spring 2026 Requests for Startups, specifically the "Cursor for Product Managers" idea by Andrew Miklas:

"Imagine a tool where you upload customer interviews and product usage data, ask 'what should we build next?', and get the outline of a new feature complete with an explanation based on customer feedback as to why this is a change worth making."

PM·AI is a working prototype of exactly that.

---

Roadmap

• Upload from URL (scrape App Store, G2, Reddit reviews)
• Notion / Linear export integration
• Product context memory across sessions
• Compare analyses side by side
• Team sharing via URL

---

License

MIT