How Nudgy handles your documents.

Documents reach Nudgy one of two ways: uploaded directly, or picked from OneDrive / SharePoint through a Microsoft 365 connection that uses a single read-only Microsoft Graph scope, Files.Read.All, plus standard Microsoft sign-in. Nudgy retrieves only the files you explicitly pick.

Nudgy never writes to, moves, renames, re-permissions, or deletes customer files. There is no code path that does.

Documents you add to a project are stored in private, access-controlled buckets, isolated per organization. When an audit runs, an isolated worker reads the stored document, extracts the schedule tables, runs the audit, and writes the findings.

Stored documents persist as part of the audit trail — in this industry, a finding is only defensible if the exact document version it was raised against can be re-opened. Your source files in Microsoft 365 are never modified; Nudgy works from its own stored copy.

Data is separated per organization and per project, enforced server-side on every query. No customer's spec index, findings, or records are reachable from another customer's context.

All traffic is encrypted in transit with TLS. Data at rest — the spec index, findings, audit records, and stored tokens — is encrypted by the managed services that hold it.

Authentication happens on Microsoft's side; Nudgy never sees or stores passwords. The read-only Microsoft access token lives in your own browser session — it is not stored on Nudgy's servers, it expires automatically, and access is revocable at any time from your Microsoft admin center.

The services Nudgy runs on, and what each one does:

Audit analysis and embeddings run on commercial API agreements with Anthropic and Voyage AI — the enterprise API terms, not consumer AI products. Your documents are sent for processing, not contributed to anything.

The current data-use and retention terms of both providers are available on request, alongside our own documentation.