Your documents become intelligence — the moment they land

Upload a 200-page policy PDF. Within seconds it’s extracted, split into ~512-word chunks, embedded with 1,536-dimensional vectors, and indexed for AI. From that moment, the policy is searchable, answerable, and ready to be used by any AI agent.

• 9 file formats indexed automatically — PDF, DOCX, CSV, JSON, XML, HTML, Markdown, and more
• Per-version status visible at a glance: queued → indexing → indexed
• Re-indexed on every new version — your AI knowledge is always current

Failure transparency built in — if something can’t be indexed, you see exactly why and can retry with one click.

Find what you mean — not just what you typed

Your team types “how do we handle vendor onboarding.” A keyword tool returns 200 documents that contain those words. NexDok returns the three documents that actually describe your process — even when the official wording is “third-party supplier qualification procedure.”

• One query searches the entire workspace — with on-the-fly filters by collection, type, zone
• Tunable similarity threshold — 0.7+ means strong relevance, lower for broader recall
• Always permission-aware — users only see content they’re authorized to see

Every result shows the matching snippet, not just the title — so you know why it’s there.

Plain-English questions, grounded answers, every claim cited

A sales rep types “What’s our cancellation policy for annual plans?” Ask returns a clear, two-sentence answer citing “Customer Agreement v4.1, Section 8.3” and “Refund Policy v2.” They click the citation, land on the exact paragraph, and ship an accurate quote in 30 seconds.

• Ask works across your entire authorized corpus
• Deep Ask scopes to one document — summarize, extract, compare
• Every claim links to source — no untraceable hallucinations

Configurable similarity threshold flags low-confidence answers instead of faking them.

Documents that matter shouldn’t go live by accident

Marketing drafts a customer announcement and submits for approval. Compliance, Legal, and the CMO each get an inbox entry. Within an hour all three approvals are in — the version publishes with a clean audit trail showing who approved, when, and what they said.

• Configure approvals per document type, per collection, or per document
• My Approvals inbox surfaces exactly what’s waiting on each approver
• Every decision logged: identity, timestamp, comment, exact version

Tamper-evident audit trail. Exportable. Webhook-deliverable to your compliance dashboard.

Every change preserved, every revision recoverable

A policy goes through 7 revisions, 3 approval cycles, and a final publish. A year later, a regulator asks “What did this policy say on March 15?” You answer in 5 seconds — and prove it.

• Sequential numbering: v1, v2, v3 — automatic, never out of order
• Head and Published pointers — always one official version, no “Final_FINAL_v3”
• Side-by-side comparison, branching from history, friendly labels

Mix native-editor versions and binary uploads in the same history — one story per document.

Permissions that scale — from 5 people to a regulated enterprise

A new analyst joins. They get a Viewer role and Editor access to the Marketing collection. They cannot see, search, or even know about the existence of the 400 confidential HR documents. The AI respects every permission, every query.

• Workspace roles + Access Groups + Security Zones + per-document ACLs
• Tenant isolation across application, database (row-level security), and storage
• MFA optional for all, required for admins — via standard authenticator apps

Manage by group, not by user. Onboarding a new HR hire? Add to the HR group — they instantly inherit every permission.