canAIbuild

alternative · reviewed · reviewed 2026-09-24

Can AI build a read-later app like Readwise Reader?

YES, BUT

The short answer

Yes, but. For personal use, AI can build a read-later app that replaces Readwise Reader ($9.99/mo billed annually) as a link list in 2–3 hours. The biggest limitation is saving full article text: browsers block cross-site fetching, so you need a small proxy or paste text in. Start by exporting your Reader library as CSV.

Difficulty
Intermediate
Build time
2–3 hours for a link list, 10–15 hours with saved article text and highlights
Build cost
$0–$20
Ongoing cost
$0 on a free proxy tier; [VERIFY] limits

AI can handle

  • A saved-links list with Inbox, Later, and Archive
  • Tags, notes, and search across titles and saved text
  • Highlights on pasted or fetched text, exported as Markdown

The hard parts

  • Fetching article text, which browsers block without a proxy (CORS)
  • Saving from your phone's share sheet into a web app
  • Storing hundreds of full articles without hitting localStorage limits

Build this first

The sensible first version

  • Save a link with title and tags
  • Inbox, Later, and Archive states
  • Reader view for saved text
  • Highlights exported as Markdown
  • CSV import from Pocket or Readwise exports

Ready to build

A starter prompt for this

Paste this into ChatGPT, Claude, or Cursor to get a working first version. It bakes in the scoping decisions above so you don't have to re-derive them.

Build a personal read-later web app to replace Readwise Reader for my own use.

Tech rules:
- Plain HTML, CSS, and vanilla JavaScript in one file. No framework, build step, account, or server.
- Store the item list in localStorage. Store article text and highlights in IndexedDB, because full text quickly exceeds localStorage limits.

Data model:
- Item: id, url, title, siteName, tags [], status (inbox | later | archive), savedAt, readAt, note, hasText.
- ArticleText (IndexedDB): itemId, html (sanitized), wordCount.
- Highlight (IndexedDB): id, itemId, text, note, createdAt.

Features:
- "Save" form: paste a URL, title auto-filled from the URL's domain until edited. Optional "Paste article text" box.
- Lists: Inbox, Later, Archive, and by tag. Search across titles, notes, and saved text.
- Reader view with adjustable font size, dark mode, and estimated reading time.
- Select text in the reader view to save a highlight with an optional note.
- "Export highlights" downloads one Markdown file per article.
- Accept ?url= and ?title= query parameters so a bookmarklet or iOS Shortcut can save a link by opening the app.

Edge cases the build must handle correctly:
1. Saving a URL that already exists shows the existing item instead of creating a duplicate. Treat URLs as equal after removing tracking parameters (utm_*, fbclid) and trailing slashes.
2. Pasted HTML is sanitized: strip scripts, iframes, event handlers, and styles before storing or rendering.
3. Highlights still display if the same passage appears twice in the article.
4. Deleting an item also deletes its text and highlights after one confirmation.
5. If IndexedDB is unavailable (some private browsing modes), the app still saves links and shows a notice that text can't be stored.
6. Very long articles (over 50,000 words) render without freezing the page.

Import and backup:
- "Import CSV" accepts Pocket and Readwise exports: find the URL column (url, URL, or Url) and an optional title column. Show counts of imported, duplicate, and skipped rows.
- One-click "Export JSON" (all items, text, and highlights) and "Export CSV" (items only). "Import JSON" restores after confirmation.

Optional, clearly separated:
- Provide a second file, worker.js, for Cloudflare Workers that fetches a URL and returns the page title and main article HTML. The app calls it only if I paste the worker URL into settings. The app must work fully without it.

Layout:
- Mobile-first, comfortable reading at 375px wide.

Exclusions:
- No publishing, App Store submission, accounts, payments, sharing, recommendations, or social features.

Write a short README explaining where data is stored, how to export and restore it, how to set up the bookmarklet or iOS Shortcut, and how to deploy the optional worker.

Keep the app under 1,000 lines and the worker under 100 lines. This is a personal tool, not a product.

What is the short answer?

Yes, but. AI can build a read-later app that replaces Readwise Reader at $9.99 per month billed annually or $12.99 monthly for one person’s reading list. A private list of saved links with tags and an archive takes 2–3 hours.

The “but” is article text. Reader fetches and cleans every article you save. A web page running in your browser can’t download other websites’ pages because of browser security rules (CORS). You either paste text in yourself, or deploy a tiny proxy on a free serverless tier. That step turns a Beginner build into an Intermediate one.

Option Best for Upfront cost Ongoing cost Main limitation
Build a rough prototype People who save links and read them in the browser later $0 and 2–3 hours $0 No saved text, so no offline reading
Build a daily-use personal version Readers who want saved text, highlights, and Markdown export without a subscription $0–$20 and 10–15 hours $0 on a free proxy tier [VERIFY limits] You deploy and maintain a small proxy
Keep paying for Readwise Reader Heavy readers who use PDFs, newsletters, text-to-speech, or Obsidian sync $0 $9.99/mo annually or $12.99/mo $119.88 per year for features you might not use

The ready-to-paste starter prompt at the end of this page builds the daily-use version, with the proxy as an optional add-on.

What should the first version include?

The first version needs fast saving and a clear Inbox. If saving a link takes more than 2 taps, you’ll stop using it.

  • Save a link with title and tags
  • Inbox, Later, and Archive states
  • A reader view for any text you paste or fetch
  • Highlights exported as Markdown
  • CSV import from Pocket or Readwise exports
  • Deliberate exclusion: no PDFs, EPUBs, newsletters, RSS, or text-to-speech. Those are the reasons Reader costs what it does.

What can AI build reliably?

ChatGPT and Claude build the list side of this app reliably: saving, tagging, filtering, search, and archive states. A reader view with font controls and dark mode also comes out well on the first try.

Concrete things AI gets right:

  • Duplicate detection that ignores tracking parameters like utm_source
  • A bookmarklet that opens your app with the current page’s URL and title
  • Highlight capture from selected text
  • Markdown export formatted for Obsidian or Notion
  • A 50-line serverless function that fetches an article and extracts the main text

Where will AI need human help?

The proxy needs you. AI writes it in minutes, but you create the free account, deploy it, and paste its URL into the app. If you skip the proxy, you paste article text by hand, which works for 3 articles a week and fails at 30.

Saving from your phone is the second decision. Web apps can’t appear in the iPhone share sheet the way Reader does [VERIFY for your device]. The workaround is an iOS Shortcut or bookmarklet that opens your app with the link attached. Test it on your actual phone before you cancel.

Other checks:

  • Sanitize pasted HTML so a saved page can’t run scripts inside your app
  • Keep article text in IndexedDB, not localStorage, or you’ll hit the browser’s size limit after a few dozen long articles
  • Paywalled articles fetch as a paywall page; the proxy can’t log in for you

Your reading list stays private in your browser. The proxy only sees the URLs you ask it to fetch.

Can I get my data out of Readwise Reader?

Yes. Reader offers several export options: a CSV of all your saved documents (from the command palette or your account page), an OPML file of your RSS feeds, and your highlights as CSV or Markdown from the Readwise export page.

The trap: the document CSV is a list of links and metadata. It doesn’t carry the cleaned article text [VERIFY]. After import, your build has every URL but no saved text until you fetch or paste it again. Export highlights separately, or you lose them.

Coming from Pocket? Mozilla shut Pocket down on July 8, 2025 and kept the export page live only until November 12, 2025. If you downloaded your Pocket CSV in time, the starter prompt’s importer reads its URL column. If you didn’t, that data is gone.

How long will it take and what will it cost?

Rough prototype: 2–3 hours and $0. Saved links, tags, archive, and CSV import.

Dependable daily-use version: 10–15 hours, including the reader view, highlights, IndexedDB storage, the phone saving workflow, and the optional proxy. $0–$20 for one optional AI month.

Readwise costs $119.88 per year upfront on the annual plan. A $20 AI month equals about 2 months of that. If you only need the link list, even the rough prototype replaces the subscription.

These are estimates from reviewing the feature set, not measurements from a completed build.

Which AI tool or approach should I use?

Claude is a good default for the app itself, and it explains the CORS problem clearly if you hit it.

ChatGPT is equally good at writing the serverless proxy and walking you through deploying it.

Cursor helps most here of any build on this site, because you’re managing 2 files (the app and the worker) and want to edit them separately.

Lovable can host both pieces for you, but adds an account and database you don’t need for a private reading list.

What will I need to maintain?

Without the proxy: nothing but backups. With the proxy: one free serverless account.

Recurring work:

  • Export a JSON backup monthly; it includes text and highlights
  • Check the proxy still works if sites start returning errors
  • Watch free-tier limits on the proxy host [VERIFY current limits]

Some sites block automated fetching. When that happens, paste the text by hand for that article.

Should I build it, buy it, or reduce scope?

Build it if you save web articles, read them in a clean view, and want highlights in Markdown.

Buy it if you read PDFs, EPUBs, newsletters, or YouTube transcripts in one place, or rely on text-to-speech and automatic Obsidian sync. Rebuilding that is months of work.

Reduce scope if you mostly save links and read them in the browser anyway. Build the 2–3 hour link list and skip the proxy entirely. Free tiers from Instapaper and Raindrop.io also cover this use.

Recommendation: build the link-list version first and use it for 2 weeks. Add the proxy only if you find yourself missing saved text. Keep Readwise if highlights flowing into your notes app is the reason you pay; that sync is its real product. If you journal about what you read, the Day One build pairs well with the Markdown export.

What if I wanted to ship this to other people?

Shipping this means a hosted fetch-and-parse service, accounts, sync across devices, native share-sheet apps, and handling other people’s copyrighted content at scale. It’s a large product, not a larger version of the personal build.

Starter build specification

  • User: You, for your own use.
  • Problem: Paying $9.99–$12.99 per month for Readwise Reader to save and read web articles.
  • Core workflow: Save a link from your phone, read it later in a clean view, highlight a line, export highlights as Markdown.
  • Required features: Save with tags, Inbox/Later/Archive, search, reader view, highlights, CSV import, Markdown export.
  • Deliberate exclusions: No accounts, no multi-user support, no PDF or EPUB reading, no store submission.
  • Data ownership: Item list in localStorage, text and highlights in IndexedDB. Exported as JSON, CSV, and Markdown. Imported from Pocket and Readwise CSVs.
  • Definition of done: You import your Reader CSV, save a new link from your phone in under 5 seconds, read it in the reader view, and export a highlight to a Markdown file that opens correctly in your notes app.

New to building? Read can AI build an app for me. For a reading streak, add a daily item to the habit tracker build.

Sources and verification

This app was not built and tested for this page. The verdict is based on Readwise and Mozilla documentation and a review of the feature set. Readwise’s price was checked on September 24, 2026.

Reviewed on September 24, 2026.

Frequently asked questions

How much does Readwise Reader cost?

Reader is included in the full Readwise plan at $9.99 per month billed annually or $12.99 billed monthly. The $5.59 Readwise Lite plan doesn't include Reader. New users get a 30-day trial.

Can I export my data from Readwise Reader?

Yes. Reader exports a CSV of all saved documents, an OPML file of your feeds, and your highlights as CSV or Markdown from the Readwise export page.

Can I still export my Pocket saves?

No. Pocket shut down on July 8, 2025, and Mozilla kept the export page open only until November 12, 2025. If you downloaded the CSV before then, it imports into any tool that reads a URL column.

Can ChatGPT or Claude build a read-later app?

Yes. Both build the list, tags, and reader view quickly. Fetching article text from other websites needs a small server-side proxy, which AI can also write, but you deploy it yourself.

Is there a free alternative to Readwise Reader?

Instapaper and Raindrop.io both offer free tiers for saving links. A self-built list costs $0 if you skip full-text fetching.