How KLARMASS works
From an uploaded document to a documented decision — in six steps, each of which stays inspectable.
1 — Upload and parsing
PDF and DOCX files are parsed server-side. The original file is stored unchanged and every extracted passage keeps a reference to its position in the original.
2 — Claim discovery
KLARMASS identifies externally checkable propositions and records the claim type. Statements that cannot be checked outside the document are not presented as if they had been checked.
3 — Evidence inspection
For each claim, the inspection uses only the parsed content and the in-document references of that single document version. It records evidence class, provenance and what is missing.
4 — Verification
Discovery finds candidate sources; KLARMASS then retrieves and reads them server-side and stores the passages that actually speak to the claim. Every source stays visible with its origin.
5 — Resolution
Each finding is resolved by a human: source added, wording adjusted, treated as an assumption, or consciously accepted. The record keeps who decided what, and when.
6 — Review and disposition
A reviewer sees the same evidence the author saw and issues a disposition. Approval by the author of the work product is not possible.
Progressive verification
Verification is progressive and bounded, not exhaustive. Only a limited number of claims per document version are checked externally; KLARMASS verifies where verification is useful and possible, and shows how much of the document that covers. Where a source cannot be reached or read, the verification stays open and is shown as open.
An inspection result describes one document version. Changing the document creates a new version, and earlier results stay attached to the version they came from.