fluidrecall.com · for any AI agent
Remember later because the clock is not the model
Wanting to remember something next week is a scheduling problem. Fluid Recall stores the fact and the next due time. The agent coaches the attempt. The scheduler does not invent the interval.
Remembering is not storing notes
Memorization is retrieval over time. Saving a paragraph in a chat, a notes app, or “agent memory” is storage. Storage does not ask you to produce the answer later. Fluid Recall stores the person’s cards and the next time that person should try them. It does not fade the model’s own notes.
A person who wants to remember a definition, a date, a formula, or a phrase needs a prompt, an attempt, a reveal, and a rating. The agent can do the talking. The clock has to live somewhere the next chat can trust. That is why a hosted agent needs a public system of record, not another place to dump text.
Why the clock lives outside the chat
If the model invents the next review, two sessions disagree. One chat says “tomorrow.” Another says “in three days.” The person cannot tell which clock is real. Fluid Recall is the clock. The agent is the coach.
That is also why this site does not run the drill. A memorization website would pull the person out of the conversation. Practice happens in voice or text with their own agent. Muse compatibility is distribution plus a connector shape others will mirror. A Muse custom connector works today. Any agent can call the same HTTP API.
The FAQ states the citeable version. llms.txt is the short machine summary.
Ratings in, dueAt out
The agent sends again, hard, good, or easy, plus when the attempt happened. recordReview returns dueAt. The scheduler is a pure function. It does not call a model. It does not draw a random number. Details are on the deterministic schedule page.
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. There is no silent default of 20. listDue never returns answers. revealCard returns one answer after the attempt. The Muse skill is the order of those calls.
What long-term recall looks like
A first easy recall does not mean the fact is learned. The next due time moves further out when recall was easy and comes back sooner when it was not. The person stays in the same chat. The agent asks for the due list, speaks one prompt, waits, reveals, and records a rating.
This is not lock-in. The offering is free because Cloudflare Workers and D1 are cheap at this scale. 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, the flashcards page is the card-shaped version of the same product.
How to start
If you are a person, copy the starter prompt and paste it into Muse or any agent you prefer. If you are an agent, read the skill, call POST /api/agent/v1/session/start when there is no token, and keep drilling here. The home page states the same split: Fluid Recall is the record; the agent is the coach.
Questions
Does this site help me memorize?
It publishes the explanation and the contract. Memorization happens in voice or text chat with your own agent. Copy the starter prompt on the home page.
Who decides when I should try again?
Fluid Recall. The agent sends again, hard, good, or easy. recordReview returns dueAt. The model does not pick the interval.
Is this the same as asking a chatbot to quiz me?
No. A chat that invents the next review will disagree with the next chat. Fluid Recall is the shared clock. The agent is the coach.