A signed SAFT is twelve to twenty pages. The part you will actually reuse is eight numbers and two dates. Pulling them out by hand takes about twenty minutes per contract, which is why it usually never gets done.
Everything below appears in a standard SAFT or token side letter. If a field is missing from the contract, the honest output says "missing" rather than guessing.
| FIELD | WHERE IT HIDES |
|---|---|
| Investor legal name | Cover page and signature block often differ — use the signature block |
| Purchase amount | Section 1, sometimes a schedule; currency is not always USD |
| Valuation cap | Definitions section, not the summary page |
| Discount | Usually next to the cap; may be "greater of cap or discount" |
| Token allocation | A schedule; often expressed as tokens, not dollars |
| TGE unlock | Vesting schedule; frequently missing entirely |
| Cliff and vesting | Annex; check whether it starts at closing or at TGE |
| Pro-rata, MFN, info rights | Often in a side letter, not in the SAFT itself |
1. The summary page lies. Cover sheets are written by the other side. The definitions section controls, and it is where "Valuation Cap" actually gets defined.
2. Vesting start dates differ. One contract vests from closing, another from TGE. Put both in the same column without normalising and every downstream calculation is wrong.
3. Side letters override. An MFN clause in a side letter can change the effective cap on three other SAFTs. If you only read the SAFTs, you miss it.
4. Missing is not zero. A missing TGE unlock is not "0% unlock", it is "not stated" — and it is the single most common gap in token documents.
One deal, one row, the same columns for every contract, plus the source line each value came from. That last part matters: when a number looks wrong six months later you need to know which clause it came from.
Reading one SAFT properly is twenty minutes. Forty of them is a week you do not have. The work is mechanical, which is exactly the kind of work that should not be done by the most expensive person in the room.
TokenDeal does this extraction: upload the signed PDF, get a normalised row with a confidence note and a source line per field. The first document is free so you can check the output against your own reading before paying for anything.
Related: token warrants and side letters · building one cap table from many documents