Marginalia

A free AI reading companion for research papers, textbooks and PDFs. Highlight any passage, ask Claude to explain it, and keep every answer as a note beside the text.

• Next.js 15 (App Router) for the public site, guides and the reader
• Supabase for sign-up/login (email + Google), the library database and original PDF storage
• Bring your own key: questions go from the reader's browser straight to Anthropic with their own API key, which is kept only in their browser
• SEO: static public pages, per-page titles and descriptions, canonical URLs, sitemap, robots.txt, Open Graph image and structured data
• Google AdSense slots on the landing page, guides and library, off until you add your AdSense IDs

The reader itself lives in public/reader/ (the same code as the Marginalia artifact). components/ReaderHost.tsx gives it a document store, file storage and Claude.

Set it up (about 30 minutes)

1. Supabase (database and login)

1. Create a free project at supabase.com.
2. Open SQL Editor → New query, paste everything in supabase/schema.sql, and press Run. This creates the docs table, the private pdfs bucket and the rules that keep each person's data private.
3. Authentication → URL Configuration: set Site URL to your domain (e.g. https://marginalia.app) and add https://YOUR-DOMAIN/auth/callback and http://localhost:3000/auth/callback to Redirect URLs.
4. Optional, Google sign-in: Authentication → Providers → Google. Follow Supabase's steps to create a Google OAuth client and paste its ID and secret.
5. Project Settings → API: copy the Project URL and the anon public key.

2. Run it locally

cp .env.example .env.local   # then fill in the Supabase values
npm install
npm run dev                  # http://localhost:3000

3. Deploy on Vercel

1. Import this GitHub repo at vercel.com/new.
2. Add the variables from .env.example under Settings → Environment Variables (at minimum NEXT_PUBLIC_SITE_URL, NEXT_PUBLIC_SUPABASE_URL, NEXT_PUBLIC_SUPABASE_ANON_KEY).
3. Deploy, then add your domain under Settings → Domains. Set NEXT_PUBLIC_SITE_URL to that domain and redeploy.

4. Get found on Google

1. Add your domain at Google Search Console using the URL prefix method and the HTML tag option.
2. Copy the content="..." value into NEXT_PUBLIC_GOOGLE_SITE_VERIFICATION on Vercel, redeploy, and press Verify.
3. Under Sitemaps, submit sitemap.xml. Use URL Inspection → Request indexing on the home page to speed things up.

New guides in lib/guides.ts are added to the sitemap automatically. Regular, useful articles are the biggest lever for search traffic.

5. Ads (Google AdSense)

1. Apply at adsense.google.com with your live domain. Approval needs real content and a privacy page, both included.
2. Once approved, put your publisher ID (ca-pub-…) in NEXT_PUBLIC_ADSENSE_CLIENT, create three display ad units, put their slot IDs in the NEXT_PUBLIC_ADSENSE_SLOT_* variables, and redeploy.
3. AdSense will ask for an ads.txt file. Create public/ads.txt with the line it gives you.

Costs

Vercel and Supabase are free to start. Readers pay Anthropic for their own questions. Your only required cost is a domain name.