Features
Hachi is a study companion built around one loop: take material you already have, turn it into questions, practise them properly, and find the parts you keep getting wrong. This page is the whole of it, in more detail than the homepage manages.
It is free to use, there is no subscription, and nothing is behind a paywall. An account adds cloud sync and sharing; it is not required to build a reviewer or take a quiz.
There are two generator modes and they are opposites, which is why they are separate rather than one form with switches.
For Study Material you choose the question type and how many questions you want. The count can be 20, 50, 75, 100, or comprehensive, which covers everything important in the source rather than stopping at a number. If a reviewer comes back shorter than you asked for, you can extend it without regenerating what you already have.
Any upload containing an image is routed to a provider that can actually look at it, even when notes were pasted alongside, so a mixed upload reads both. Large photographs are downscaled on your device before they are sent, and a photograph that is already small is passed through untouched rather than re-encoded larger than it started. HEIC is listed and then refused with an explanation, because a file the picker silently hides looks like a broken app.
Generation is optional. The question editor writes a reviewer from nothing: a topic, a question, choices, the correct answer, and an explanation. It is slower than pasting in notes and much better for the questions that came out of a lecture, a marking comment, or a past paper you disagree with.
An editor saved from a shared reviewer updates that reviewer rather than creating a copy, so a correction everyone can see does not quietly split off into your private version.
A reviewer is not a quiz you take once. The same questions can be practised five different ways, and your progress is kept whichever you choose.
If you only take one suggestion from this page, take it: sit something in Exam Mode with the notes shut, then open Review Answers afterwards and work through what you missed. That pairing is what active recall actually looks like in practice.
Every reviewer you build, whether generated or written by hand, lands in one searchable library with its subject, its topic, and how many questions it holds. Search across titles and question text, so “where was that thing about the cardiac cycle” is a question the library can answer.
Edit, duplicate, rename, or delete any of them. Renaming a reviewer rewrites it from the copy on screen rather than from a summary, so a rename can never quietly take the questions with it.
None of this is uploaded anywhere. It lives on your device until you sign in, and then it syncs to your own account. There is no leaderboard and nothing is compared to other people.
Reviewers, quizzes, scores, and history all run from your device, so a bad connection costs you generation and sharing but not revision. Add Hachi to your home screen and it launches in its own window and keeps working with no signal at all. What works offline, in detail.
Signing in lets you sync across devices, add friends, and share reviewers. Sharing hands a copy to the people you share it with, which is worth knowing before you send something you might later want to delete. Groups give everyone one place to see the reviewers that belong to a module rather than a chat thread full of attachments.
When the questions of a shared reviewer change, the people using it are told, so nobody revises an old version by accident. Running a group that works covers how to avoid five people making five slightly different reviewers for the same section.
Being straight about the limits is more useful than a longer feature list. Hachi does not have spaced repetition scheduling built in, so the timing of your revisits is yours to arrange; the spaced revision guide is how. It does not check your answers against an official mark scheme unless you give it one. It does not use a leaderboard or contact anyone who used it. And nothing it generates is authoritative: treat a generated reviewer as a draft to check against your own material, which is also what the terms say.