AI-powered SBIR/grant proposal drafter with reviewer-simulation scorecard.
- User pastes a solicitation (NSF, NIH, DoD, NIST SBIR, etc.) + their project description.
- Free tier: parses the solicitation and shows review criteria + compliance checklist.
- Paid tier: generates section drafts (Specific Aims / Significance / Innovation / Approach / Commercialization) AND scores each draft against the actual review criteria as a hostile reviewer would.
The moat is the reviewer simulation. Anyone can prompt ChatGPT to write a grant section. Nobody else takes that draft and scores it 1-9 against the agency's specific review criteria with concrete weakness flags. That's what professional grant writers charge $200/hr for.
- Frontend: static HTML/CSS/JS (no framework), 2 routes (organic + paid-traffic variant)
- Backend: 4 Vercel serverless functions (Node.js,
/api/directory pattern) - AI: Anthropic Claude (Haiku 4.5 for free tier, Opus 4.7 for paid drafts/scoring)
- Payments: Stripe Payment Links with
client_reference_id+{CHECKOUT_SESSION_ID}redirect - Hosting: Vercel
- Domain: Namecheap
grantforge/
├── api/
│ ├── parse-solicitation.js # Free tier: extract criteria from solicitation
│ ├── draft-section.js # Paid: generate section draft
│ ├── score-section.js # Paid: reviewer simulation
│ ├── verify-payment.js # Stripe checkout session verification
│ └── stripe-webhook.js # Optional: webhook for fulfillment record
├── public/
│ ├── index.html # Organic landing page (full content, social proof)
│ ├── app.js # Shared client-side logic
│ ├── styles.css # Shared styles
│ └── d/
│ └── index.html # Paid-traffic landing page (single CTA, urgency)
├── package.json
├── vercel.json
├── .env.example
└── DEPLOY_GUIDE.md # Step-by-step deployment + monetization guide
- Free: solicitation parse + compliance checklist
- $49: single section (draft + score + revisions)
- $149: full pack (5 sections, all scored)
- $299/mo: unlimited (later — start with one-time)