AI contract data extraction with page-level citations
Retyping contracts into a tracker is how mistakes get made. Keep Track of Stuff reads each uploaded document — PDF, DOCX, even scans — and turns it into structured data you can search, sort, and get alerts from.
Updated
What gets extracted from every contract
Each document is read end-to-end for the fields an operator actually tracks: counterparty, agreement type, effective date, end date, renewal type (auto-renew, manual, evergreen, fixed-end), renewal term, notice period, cost with currency and frequency, a plain-English summary, and the concrete obligations the contract imposes — insurance minimums, maintenance duties, reporting requirements.
From the end date and notice period, the system computes the date that actually matters: the notice deadline, the last day you can still act before an auto-renewal locks in or an option lapses.
Every value is cited to the page it came from
Extraction you can't check is extraction you can't trust. Every extracted value carries a citation to the exact page of the source document — click it and the PDF opens right there — plus a confidence score. Values the AI wasn't sure about are visibly flagged until a human confirms them.
Verification is one click per field, or one click for everything. After that, the record is marked human-verified, and later amendments can't silently overwrite what you've confirmed — changes arrive as suggestions you approve or dismiss.
Amendments update the record, with your approval
Contracts live as families: the original plus its amendments, extensions, and renewals. When you upload a fourth amendment that changes the rent, the AI reads it against the existing record and proposes exactly what changed — old value, new value, source page. Nothing about a verified agreement changes without you clicking Apply.
Works on scans, not just clean PDFs
Text-layer PDFs are read directly; scanned documents route through visual reading automatically. Long documents are handled in sections and merged, so a 100-page lease extracts as reliably as a 5-page vendor agreement.
Frequently asked questions
How accurate is the extraction?
Strong on standard fields, and honest about the rest: every value has a confidence score and a page citation, and low-confidence values are flagged for review. You verify each contract once — a couple of minutes — instead of retyping it and hoping.
What file types can it read?
PDF, DOCX, PNG, and JPG, up to 50MB per file. Scanned documents are read visually; digital documents use their text layer for speed and precision.
Is this the same as lease abstraction?
Yes — lease abstraction is the commercial-real-estate name for exactly this: pulling a lease's key terms into structured data. Keep Track of Stuff does it for leases and every other operator contract, automatically on upload.
Keep reading
Put your contracts somewhere that reads them.
14-day free trial · No card required · From $19/mo