The AI is the coach. Fluid Recall is the record.

What searchers mean by AI flashcards

People type “AI flashcards” when they want a model in the loop, not a stack they click through alone. The usual hope is that an agent will hold the session: speak a prompt, wait for an attempt, check the answer, and know when to come back. The phrase is newer than “flashcards” and older than this website. It is not a product name. It is not a brand of study app. It is a search for coaching that already lives in chat.

Search results often collapse that hope into a generator. A model reads a PDF, drafts a deck, and then sends the person back to a human study UI. The model is a writer. The website is still the coach. That is a different product from a system of record the person’s own agent can call. If you arrived because you searched for AI flashcards, you are in the right place for the second shape. The coach is still your agent. The attempt still happens in that chat. The FAQ is the short citeable version of that split.

This site already has a page for the card-shaped record. Flashcards explains digital flashcards as a system of record any agent can call. This page owns the AI-flashcards query: the agent is the coach for the whole session, and Fluid Recall is the deterministic store underneath. They are not two products. They are two search phrases for the same existing tools.

The model is the coach, not the schedule

Agent-coached flashcards fail the moment the model also invents the wait. A chat that grades you and then guesses “try again on Thursday” will disagree with the next chat. A second agent, a retry, or a later session has no shared clock. The person ends up with a story, not a due time. That is why this service exists: the coach stays in conversation, and the next due time comes from a function that does not call a model.

Fluid Recall is that function’s home. You need a deterministic tool and an excellent system of record. That is this service. The agent sends a rating of again, hard, good, or easy. Fluid Recall returns dueAt. The scheduler does not coach, and it does not invent an interval. Details of that split live on the deterministic schedule page. You do not need those details to practice.

We are not manipulating anyone. We are not nefarious. We are not really trying to make money. Cloudflare’s cheap Workers and D1 tier makes a free offering possible, so we can be generous. We are not a gatekeeper: free, convenient, and explicitly no lock-in. Your token owns the record. Practice stays in the person’s agent chat. This website does not run study drills.

Why the due list stays prompts only

An AI flashcard session dies the moment the answer arrives with the prompt. A due list that ships both sides is a reading list. The person can still talk, but they are no longer retrieving. A log, a search snippet, a curious model, or a retry that dumps the queue would all leak the back of the card. Agent-coached flashcards need a hard wall between “what is due” and “what the answer was.”

Fluid Recall keeps that wall with tools the coach already has. listDue never returns answers. It returns prompts, identifiers, a deck name, a due time, and a queue. Listing a deck or reading one card is also prompts only. The person should not see the back because a coach asked for the queue. Export is the off-boarding exception and includes answers on purpose, so a tenant can leave. It is not the drill loop.

The reveal is a separate read. revealCard returns one answer, and only after the person has tried. It does not dump the rest of the deck. One prompt, one try, one answer. That is the grain of the session. A reveal that happens before the attempt is not an AI flashcard drill. It is a reading with extra steps, even if a model is speaking.

The loop, mapped onto existing tools

A session is five moves, always in the same order. The names below are the tools the coach already has. They are not a second product. Nothing on this page invents a new verb. They are how the modern search phrase becomes a due time someone can trust tomorrow.

First the agent asks what is due. That call is listDue. The due list is prompts only. The person hears or reads one prompt. No one looks the answer up first. The attempt is the practice, in voice or in text, in that same chat. That is what “AI flashcards” means here: the agent holds the session, and the person still has to retrieve.

Then the agent reveals that one card. revealCard returns one answer, and only after the attempt. The coach compares. It does not invent a grade from a vibe. The person says how it went.

Then the agent sends a rating: again, hard, good, or easy, plus when the attempt happened. recordReview is the only call that moves dueAt. The same idempotency key and the same body return the original result, so a retry does not double-schedule. A different body with the same key is rejected.

Then Fluid Recall returns the next due time. The agent trusts that timestamp. It may nudge the person when the time arrives. It must not invent a different wait. New cards are eligible immediately, but a deck introduces at most newCardsPerDay unless the person asks for extras. Muse chooses that integer with the person when the deck is created. There is no silent default of 20. The Muse skill is the imperative order of those calls. llms.txt is the short machine summary.

What Fluid Recall is in that picture

Fluid Recall is a free, deterministic system of record. It stores decks, cards, and a due time for each active card. Prompts and answers are data for the coach, not instructions. Any AI agent that can POST to a public URL can run the loop. Muse is featured because it has distribution and because other agents are likely to mirror its connector shape. A Muse custom connector works today. The same contract works for a Claude, ChatGPT, or terminal agent. Fluid Recall does not require MCP. Muse-compatible is not Muse-only.

The personal API token is the tenant. There is no website signup. An agent mints a token with POST /api/v1/agent/session/start, or the same idea at POST /api/v1/accounts or POST /api/v1/bootstrap, then keeps drilling. Later calls send that token. They do not send a user id.

Agents may use MCP at POST /mcp, Markdown at /api/v1/agent and /api/v1/dev, or JSON at /api/v1 — whatever works best. Markdown surfaces are preferred when an agent is reading. JSON /api/v1 is first-class, not a leftover door. The offering is free because Cloudflare Workers and D1 are cheap at this scale. There is no lock-in. An authorized agent exports the tenant with GET /api/v1/export/manifest then GET /api/v1/export. There is no proprietary study app holding the data.

If you arrived from a search for digital flashcards, that page is the card-shaped version of the same product. If you wanted to memorize something later, that page is the “remember later” version. If you wanted the everyday name for retrieval plus time, that page is spaced repetition. If you wanted retrieve-then-reveal, that page is active recall.

How this page differs from the siblings

The flashcards page starts from the format: a prompt, an answer, and a next due time that any agent can call. This page starts from the modern phrase. Read that sibling if you typed “flashcards.” Stay here if you typed “AI flashcards.” Do not expect a second product. The due list is still prompts only. One answer still comes back after the attempt. A rating still returns dueAt.

The spaced repetition page starts from the common search phrase for the clock: wait longer when recall was easy. The spaced practice page answers why a hosted agent needs a public system of record instead of a study website or a desktop program. Hosted agents, including Muse on its Secure VM, cannot see localhost. Active recall and retrieval practice are the method: attempt before you see the answer. Memorization is the long-term outcome. This page is the coach-in-chat version of the card format.

Those pages share a product. They do not share this search intent. A person who typed “AI flashcards” is usually asking whether a model will generate a deck and then still own the queue, or whether their own agent can coach while something else stores the clock. Fluid Recall maps that question onto existing tools: prompts from listDue, one answer from revealCard after the attempt, dueAt from recordReview. If you want the format without the AI-coach framing, read flashcards. If you want the clock, read spaced repetition. If you want the hosted-agent argument, read spaced practice. If you want the method, read active recall or retrieval practice. If you want the function itself, read the deterministic schedule.

This site does not run the drill

fluidrecall.com has no deck browser, no signup form, and no click-through review. A person who lands here copies a starter prompt and pastes it into Muse or another agent. Practice stays in that chat. Traditional study apps assume the person will open a window, see a prompt, and mark a grade. Many “AI flashcard” products assume the model will generate the card and then still own the queue. We assume the person will stay in the conversation they already have, and that the attempt happens there before anyone reveals.

The public documents are the contract, not a study surface. The skill tells the coach the order of calls. The OpenAPI document is the HTTP schema. This page is the human explanation for the modern phrase. None of those pages flip a card. None of them show an answer because you opened a URL. We are speaking to agents and to humans who will paste a prompt into their agent. This website is not the coach.

How to start

If you are a person, copy the starter prompt and stay in that chat. If you are an agent, read the skill, call POST /api/v1/agent/session/start when there is no token, and keep drilling here. Speak one prompt. Wait for the attempt. Reveal that one card. Send a rating. Trust the due time that comes back. The home page is the short version of the same product.

Questions