Two tools. Use them separately or together. The teacher stays in the pilot seat; the AI handles the autopilot work.
Upload your test PDF. Get a ready-to-mark Excel workbook with auto-calculated totals, conditional formatting, Question Level Analysis, and a per-strand breakdown.
Upload your filled workbook. Get a tailored, growth-oriented report comment for every student, anchored to their actual performance.
No login. No accounts. No student data stored.
English, HASS, Arts, Languages, Tech — anything you mark with a rubric. Upload the rubric, get a workbook with the rubric's actual levels as dropdowns, generate feedback in five different formats.
Upload your rubric (PDF or paste text). Get an Excel workbook with one row per student, one column per criterion, and dropdowns of the rubric's own levels in each cell.
Upload your filled rubric workbook. Pick a feedback mode: parent report, student-facing draft feedback, PTI talking points, external email, or semester synthesis.
Same workbook flows through the whole assessment cycle. No re-uploading. No re-tagging.
Claude reads every question, estimates marks, and tags each one to an AC v9 strand. You review and tweak the tagging in one screen.
One click gets you an Excel file with a marking grid, formulas for totals and percentages, class analytics, per-question difficulty, and a per-strand breakdown.
Mark however you usually do. Cells turn green for full marks, red for zero. Validation rejects out-of-range marks. Analytics auto-update as you go.
Upload the filled workbook back. Get a personalised report comment for every student, plus a flagged strategy focus. Comments are written back to the same workbook.
Built from real student data. Anchored where they scored full marks and where they didn't. No question numbers, no formulas, no judgements on the overall result.
Sam worked through compound interest problems with weekly compounding accurately and completely. The next area to develop is converting expressions with negative indices to positive index form. To build this skill, Sam would benefit from working through several examples step by step, starting with simple expressions and progressing to multi-step ones.
Jamie converted between scientific notation and standard form accurately, showing steady recognition of place value across the test. An area for development is compound interest with non-annual compounding, where the calculations were not yet attempted. A useful next step is working through step-by-step examples that break down how to adjust the interest rate and number of periods when compounding occurs more often than annually.
Two real students. Different scores. Same respectful, growth-oriented tone.
Autopilot does 95% of the flying. Pilots own the worst-case scenarios. The same architecture works here: you mark the papers, apply your judgement, run your school's marking scheme. Markpilot does the operational tail — calculating, analysing, comparing, synthesising, writing — the boring stuff that costs you a Saturday but doesn't need you to be there.
We don't grade your papers. You do. We don't store your students' data. You close the tab, they go home with you. The accountability that lives with the teacher stays with the teacher.
I teach Year 9 maths. Every reporting cycle I'd sit down on a Saturday with thirty students' tests, a marking sheet, and a Word document open for comments. Five hours later I'd realise three of the comments were near-identical and I'd lost the energy to make the next one specific.
Markpilot is what I built so I could spend Saturday with my family instead. I figured if it worked for me, it'd work for the rest of the maths department. It does the operational tail that doesn't need a human, and stays out of the part that does.
The boring stuff done properly. The smart stuff done carefully.
Strands and skills tagged automatically against the current Australian Curriculum. Edit anything before generating; nothing's locked in.
Question Level Analysis plus class mean, median, mode, std dev, pass rate, grade distribution, per-strand rollup. Formulas, not images. Updates as you mark.
No question numbers, no formulas, no padded phrases, no judgements on the overall score, no inferred internal states. Student-as-subject sentences only.
One-line style preferences apply to every comment forever. Tweak any single student's comment with a feedback note and regenerate.
Nothing persisted, no backend database, no analytics on your usage. PDFs and marks are sent only to Anthropic for processing.
The Excel file flows through the whole cycle. Marks, analytics, charts, and comments all live in the same file you download once.
No accounts. No database. No tracking. Each upload is processed by Anthropic's API and immediately discarded. We literally can't show you a list of your past tests because we don't have one.
Answers from a maths teacher who built this for their own marking pile.
Yes. The PDF is processed as both text and image, so a clean scan works as well as a digital PDF. Very low-quality scans or photographed PDFs may need a second look at the extracted tags before generating.
Built specifically for secondary maths in Australia (AC v9). Other quantitative subjects may work but the strand tagging is hardcoded to the maths strands (Number, Algebra, Measurement, Space, Statistics, Probability).
Yes. After extraction you get an editable grid. Change the question label, max marks, strand, or skill before downloading the workbook. If extraction misses a question, add it manually.
Two modes. Per-question (default) combines sub-parts into one row (Q4 = 4a + 4b + 4c with summed marks). Per sub-question keeps them separate. Multiple choice sections can be a single row or one row per item.
Good enough that they need light editing rather than a full rewrite. Comments are anchored to the actual data (full marks question = strength, zero-mark question = gap), follow strict tone rules (no padded teacher-comment cliches, no question numbers, no judgements on overall score), and are written student-as-subject for a professional flow. You can tweak any single comment with a one-line feedback note and regenerate.
Nothing persists. No backend, no database, no analytics. Your PDF and marks are sent to Anthropic's API for processing and discarded immediately. Your API key (if you use the BYOK free version) is stored only in your browser. No accounts, no tracking. We literally couldn't show you your usage history because we don't keep one.
On the roadmap. The current comment format is parent-facing report style. Student-facing feedback and parent-teacher interview talking points are next on the build list.
A secondary maths teacher who got tired of spending hours every reporting cycle writing variations of the same comment for thirty students at a time. Built for personal use first, then opened up.