Short answers are read by a trained AI model against your rubric — not a district-locked black box. Student work is processed to produce a grade and isn't retained afterward or used to train any model.
No answer-key spreadsheet, no tab names that have to match exactly. Paste the questions and AI drafts the rubric and the key points that earn credit.
Multiple choice is matched instantly. Short answers are read against your key points and get feedback a student can actually act on.
Weak and borderline answers are flagged for your eyes. Nudge any score, then publish to the Classroom gradebook as drafts you still control.
Placeholder quotes — swap in real ones from your pilot teachers before launch.
"I graded a unit quiz on my prep period instead of my kitchen table. The feedback was better than what I write at 10pm."
"The flagged-for-review pile is the whole product. It shows me the eight answers that actually need a human."
The real extension popup, at real size, with sample student work. Tap the steps to move through a full grading run.
Not a black box. Here's exactly where your Google data goes, where it doesn't, and what a trained AI model ever sees.
Tally signs in with your own Google account and calls Classroom directly from the extension. Rosters, names, and submission IDs never leave your browser.
Each answer is paired with a random id generated on your machine — not a student id. Name, email, and course are stripped before anything is sent anywhere.
Tally's server holds the model key and reads only the rubric plus the anonymized answer text. It grades, replies, and never stores the answer.
The random id comes back matched to a score and feedback. Only your browser knows which student that id belongs to — the server never did.
Tally's own servers never receive your Google account, your course information, or a student's identity — only the anonymized text of an answer, for the moment it takes to grade it.
That boundary is enforced in code, not just policy: the grading server rejects any request that looks like it contains a name, email, or Classroom ID before it processes anything.
No name in. No name out. Ever.
Three minutes to set up, and the free tier needs no API key. Bring a department and everyone shares the same rubric library.