Anchor Cards

Anchor Cards is a spaced repetition app you work through with an AI agent. The agent writes the material; you review it; the app does nothing but quiz you. It ships with no built-in cards. Your cards are Markdown files, kept in your own GitHub repository or in Anchor Cloud (no sign-up). The app reads them, quizzes you, and writes back only your study log. Cards never live inside the app, so you can always edit them in your usual editor. * It teaches, you review Just tell your AI agent what you want to study, and it writes the cards on the spot and puts them where you keep them. The app has no screen for creating cards — with no path to create one, there's nothing for the app itself to break. Material is never finished when it is written. Spot a wrong explanation while reviewing, have the agent fix it, and go back to reviewing the corrected card. That loop runs without ever closing the app. * Your Markdown is the source of truth Write two lines at the top of a Markdown file — id and deck — and it becomes reviewable. Put the file anywhere, name it anything. Deck names and headings are whatever you wrote. No export feature is needed, because the files were already Markdown. Uninstall the app and your cards stay exactly where they were. * Stop copying the same definition onto ten cards Copy one definition onto ten cards, and every time the definition changes you have to fix all ten — and one of them always stays stale. In Anchor Cards, the full text lives on a single definition card. Other cards only reference it. At review time, the app inlines the referenced card's back into the one you're reading, so it still reads as a single card. The stored text never changes; you only ever edit the one definition card. * Hand a card straight to your AI agent Mid-review, if an explanation looks off, copy the card's stable ID with one tap — that alone is enough for your agent to find it in the same repository. Cards opened from the Search tab copy just their ID too. Every sync also writes a ranking of your weak cards back to the repository, so your agent can see what you're struggling with just by opening the repo. * Read the fix inside the app When your agent proposes the fix as a pull request, the Changes tab renders that card exactly as it will be quizzed — not a diff, but the corrected text itself, so you can decide whether to merge right there. Leave it open and it updates as new commits arrive. * Review intervals from FSRS Anchor Cards schedules reviews with FSRS (Free Spaced Repetition Scheduler). Cards you know well come back less often; shaky ones come back sooner. * Getting started On first launch, the app runs on three bundled sample cards, enough to try the whole flow. Keep your own cards in one of two places: - Anchor Cloud: create a vault in Settings and give the line it shows to your AI agent. - GitHub: go to Settings -> GitHub connection, sign in (or paste a personal access token), pick a repository, and sync. Your Git history becomes your study history. No account required. If you want to connect other devices or apps, you can optionally sign in with your email. * What the app talks to Your GitHub access token and Anchor Cloud key are stored only in the device's encrypted storage (Android Keystore). Cards sync only to the GitHub repository or Anchor Cloud you chose. If you sign in with email, we never store the address itself — only a hash used to match it. The app talks to Google Play and our billing backend (RevenueCat) only to confirm an Anchor Pro purchase, and it does not track your behavior. * Anchor Pro (optional, one-time purchase) Every feature works for free. The only difference is how many new cards you can start per day (20 for free). Anchor Pro removes that limit instantly. It's a one-time purchase, not a subscription. * What you need - A place for your cards: Anchor Cloud (no sign-up) or a GitHub repository (private is fine) - One card per file (put it anywhere, name it anything)