How Bento uses AI

What the AI does, what it never does, and where your data goes. Last updated July 22, 2026.

On Bento Plus, the AI is Claude, made by Anthropic. On the buy-once plan, you choose Anthropic or OpenAI and use your own API key; a Custom option accepts any other OpenAI-compatible service, untested and marked as such in the app. If anything here ever stops being true, this page changes first.

What the AI does

• Organizes syllabi. Drop in a syllabus file and it extracts the instructor, office hours, grading scheme, dates, and topics into the class notebook, with the original file kept so you can always check its work.
• Builds study guides from your own notes. It reads the notes you wrote for a class and organizes them into concepts to review. When it believes the exam will cover something that isn’t in the notes, it says so explicitly: the gap is labeled as the AI’s suggestion, with the reason it flagged it, and you decide to keep or dismiss it.
• Writes practice questions and grades the attempts. Questions are tied to chapters; you answer in your own words and get honest “got it / partially / revisit” feedback per question.
• Reads returned, marked work. Upload a graded test or assignment and it summarizes where marks were lost, why, and what to change, including noticing when the same mistake has cost marks twice.
• Builds a game plan for the marks. Given the grading scheme and results so far, it lays out where the remaining points are and what each one costs to earn.

What it never does

• Bento won’t write essays, complete problem sets, or produce assignment answers on your behalf. Its AI features organize, quiz, explain, and plan.
• What you write stays exactly as you wrote it. AI output lives in its own labeled space and never overwrites your notes.
• Marks come from the syllabus and your own records. Anything the AI adds beyond that is clearly marked as its suggestion, not fact.
• AI runs only when you press a button, never on its own, and nothing is sent anywhere in the background.

Where the data goes

• Buy-once (your own key): when an AI feature runs, the relevant material (the syllabus text, the notes being turned into a guide, the uploaded marked test) goes from your computer directly to the API of the provider you chose, under your own account and that provider’s privacy terms (Anthropic · OpenAI). If you configured a Custom endpoint, the same material goes to that service instead, under whatever terms you have with it. Bento has no server in the loop, and we never see anything.
• Bento Plus: the same material passes through our server, which exists only to attach our AI credentials and forward the request to Anthropic, which handles it under its own API data-retention terms. We do not retain or read the content of what passes through; what we keep is token and request counts per license (to run the monthly allowance). Details, including Anthropic’s retention, in the privacy policy.
• Everything else stays home. The planner, notes, marks, and every generated guide are stored on your computer, not with us.

The models, named

Bento Plus runs on Claude models from Anthropic. The buy-once plan runs on whichever provider you choose in Settings: Claude models from Anthropic, or models available through the OpenAI API. Those are the ones we test Bento’s features against. A Custom option accepts any other OpenAI-compatible service, but those aren’t quality-checked, and the app says so right where you set it up. Whichever you use, Bento’s AI features must be used within that provider’s usage policies. That obligation passes through to you, and it’s echoed in our terms.

Why this page exists

Because “AI-powered study tool” can mean anything from a tutor to a plagiarism machine. Bento was originally built by a parent for one university student. The AI was designed around a simple constraint: it should help him study, not quietly do the studying for him.