Words in context, the way the test asks

Because that is how the test asks. The current digital SAT tests vocabulary as "Words in Context": you get a sentence and have to work out which word fits from the surrounding logic. It does not ask for definitions of obscure words the way the old test did. Memorising a definition list trains the wrong skill — you need practice reading a sentence and noticing the clue that rules three options out.

Because that is where the learning is. On a real question at least one wrong option is genuinely tempting — a near-synonym with the wrong connotation, or a word that fits the grammar but not the logic. Being told "correct" teaches nothing. Being told why the tempting option was tempting is what transfers to the next question. Every item here explains all three distractors, and the tests enforce that.

Not on a server — there is no account and nothing to sign into. Every practice set has a seed, shown with the results, and the same seed always returns the same questions. So you can bookmark or share a set and come back to exactly it. That is deliberately simple: no account means nothing to lose access to and no data of yours to leak.

Yes, if generation is switched on for this deployment. Paste a passage — an article, a chapter, or the questions you got wrong on a practice test — and it will write items testing words that actually appear in it. You choose the model, and the cheaper and better options are priced differently because they genuinely differ at the hard part: writing distractors that are tempting rather than obviously wrong.

No, and the tool says so on every generated response. Nobody has reviewed them. The common failure is a question with two defensible answers, which is worse practice than no question at all because it teaches you to distrust your own reasoning. Use generation for coverage of your own material, and the curated bank when you want to trust the answer.

Foundation items put the context clue close to the blank and point one way. Core items sit where most of the test does — at least one distractor is genuinely plausible until you find the clue. Stretch items use near-synonyms that differ only in connotation or register, which is the hardest version of this question type.

No. Not affiliated with, endorsed by, or connected to the College Board, and these are not official practice questions. SAT is their trademark; it is used here only to say what this practises. For official material, go to the College Board directly.

That is part of why it exists. There is an MCP server at /api/mcp with three tools: draw_items to get questions, check_answer to mark an attempt and get the teaching content, and generate_items for a learner's own material. The draw response deliberately contains no answers, so an agent can present questions without leaking them. There is also a JSON API at /api/v1/ and an OpenAPI document at /.well-known/openapi.json.