AnswerIt AI
Context
A solo side project that grew into a real product. AnswerIt AI lives in the browser and closes the gap between "I have a question about this page" and "I have an answer" — no copy-paste, no tab switching, no prompt engineering.
Problem
LLMs are great at answering questions, but the friction to get a good answer from a web page is real: copy the text, open ChatGPT, paste it, re-add the context, wait. For quick look-ups that flow breaks focus entirely. Diagrams and screenshots make it worse — they're awkward to paste into text-only tools, and the model needs the surrounding context to be useful.
Approach
- Capture UX: a snipping tool built as an MV3 content script with a Svelte overlay. Select any region — text, image, or a mix — and send it through to the answer pipeline.
- AI pipeline: routed through n8n so I can swap models and iterate on the prompt chain without redeploying the extension.
- Backend: Supabase handles auth, usage metering, and per-user history.
- Monetization: Lemon Squeezy for subscription checkout — paid tier launched without wiring up Stripe + VAT compliance from scratch.
Outcome
Shipped at answerit.pro with a working free tier and a paid plan. The MV3 + Svelte shell I built for it became the reusable base for a smaller companion extension (Youtubely).
Stack notes
- Manifest V3 + Svelte for a tiny, reactive content script surface
- Supabase for auth, DB, and usage tracking
- n8n as the orchestration layer for the AI pipeline
- Lemon Squeezy for checkout and tax compliance
- TypeScript everywhere the extension talks to itself