Hi! Great questions. Here’s the quick breakdown:

RAG: Yes. LetterMint includes a built-in Knowledge Base / Document Q&A system. Upload .txt, .md, .csv, or .json files and the AI answers using that content locally in-browser (no cloud). Current version uses chunked context injection with configurable chunk size/overlap. WebNN vector search is planned for v2.0.

Custom Skills / Prompts: Yes. The template includes 30+ tools with dedicated system prompts. You can also create and save custom prompts/templates as reusable AI “skills.”

MCP (Model Context Protocol): Not built in. MCP support isn’t included out of the box. Since everything runs locally via WebGPU, external MCP servers would require custom backend integration.

Model Training / Fine-tuning: No. The app runs pre-trained local models like Llama 3.2, Phi 3.5, and Gemma 2. Browser-based retraining isn’t supported, but prompts + RAG context give strong customization without fine-tuning.

Hope that helps! Let me know if you have more questions.

Thank you!

Hi thank you so much for the kind words, great questions – and yes, this product has real potential for the model you’re describing!

Extended License ($399): The Envato Extended License allows you to use LetterMint as the foundation of a paid end-product – a SaaS, hosted app, or subscription service where your users are paying customers. Regular License covers free/internal use only; Extended is the correct choice for monetization.

How it works as a SaaS: The core AI runs 100% in the browser via WebLLM + WebGPU – no per-request API costs, no backend inference server. Your users’ devices handle all AI processing. You deploy once to a static host (Vercel, Netlify, Firebase) at near-zero infrastructure cost.

For a full multi-user SaaS, you’d integrate:

Auth – the AuthContext is already abstracted; plug in Firebase, Supabase, or Clerk

Cloud storage – swap localStorage for Firestore/Supabase (documented in the code)

Billing – connect Stripe/Paddle to the existing “Upgrade Plan” UI hook

Lifetime subscription angle: Since inference costs are zero (runs on the subscriber’s device, not your server), your operating cost doesn’t scale with usage – which makes the lifetime deal math genuinely work here, unlike cloud-API-based tools.

The demo appears simple because it runs standalone without a backend. The integration hooks are built in and documented – they’re just not wired to a live backend in the demo.

Happy to answer any follow-up before you decide!

Thank you!

follow the instructions in the readme file… it works. good stuff.

Hi eliaseg, glad you got it working!

That’s exactly it – the README and Docs/ folder walk through everything step by step. Thanks for confirming it for other buyers who might hit the same thing.

Enjoy LetterMint!

Thank you!

Hi Raju, thanks for reaching out!

A blank page on shared hosting is almost always one of two things:

1. The dist/ build folder wasn’t uploaded LetterMint is a React app – you need to build it first, then upload only the output files.

npm install npm run build This generates a dist/ folder. Upload the contents of dist/ (not the source folder) to your hosting’s public_html directory.

2. Missing .htaccess for SPA routing React Router requires all URLs to redirect to index.html. Create a .htaccess file in your public_html with:

apache <IfModule mod_rewrite.c> RewriteEngine On RewriteBase / RewriteRule ^index\.html$ – [L] RewriteCond %{REQUEST_FILENAME} !-f RewriteCond %{REQUEST_FILENAME} !-d RewriteRule . /index.html [L] </IfModule> Also note: LetterMint requires Chrome 113+ or Edge 113+ for full AI functionality (WebGPU). On other browsers it runs in Fallback Mode, which still works fine.

Please try the above steps and let me know if you’re still seeing the blank page – happy to help further!

here the demo Link: https://preview.codecanyon.net/item/lettermint-ai-browsernative-ai-content-writer-react-app/full_screen_preview/63146970

click Login with Demo Account button below