Callback
Talking to Claude is a remarkably good way to learn something. It's a terrible way to remember it. Callback closes the loop: connect it in Claude, and at the end of a learning conversation it saves what you covered as spaced-repetition cards — then your phone pings you with each one right before you'd forget it, so what you learned this week is still yours next year.
$ curl -fsSL https://callback.cards/install.sh | bash
How it works
Add the connector in claude.ai
In claude.ai → Settings → Connectors → Add custom connector, paste https://api.callback.cards/mcp. See the full walkthrough →
Sign in or create your account
When Claude prompts you, connect your Callback account. It's invite-only right now, so you'll need to be on the allowlist.
Chat with Claude, then keep it
Learn the way you already do — any topic, any depth. At the end of the conversation, Claude distills what you worked through into a handful of well-formed cards and saves them to your account.
Review in your browser or on your phone
The moment you have cards, review them right in your browser at callback.cards/app — nothing to install. Want a nudge when they come due? The Android app pings you — miss a card and it returns sooner, master it and it stretches out, each review timed to land just before the memory fades.
Why it's different
The cards write themselves
The classic reason people abandon flashcards is authoring. Here the tutor who just taught you writes the questions — from the conversation you actually had, not a generic deck.
Real recall, really graded
Typing an answer from memory beats recognizing one in a list — but ordinary apps can't grade prose. Claude grades yours semantically, so free response is the default, not the fallback.
The tutor stays in the loop
A question you keep missing doesn't just keep coming back. It goes back to Claude, which re-teaches the idea from a fresh angle and writes a better question. Failure is a teaching problem, not a scheduling problem.