Compare deployment modes
Three deployment modes, one verification gate. The difference is who runs it and where. Here is the whole decision on one page.
The three modes at a glance
Two meanings of zero-egress — never blurred
Both are real; they are not the same thing:
- On-prem (Mode 1): zero-egress = as installed, DocMark sends nothing off the box — no model call, no DocMark telemetry. It runs offline once installed; keeping the box off the internet is your network’s control.
- DocMark cloud (Mode 3): zero-egress = no PII reaches the execution plane — your identity stays in the control plane; the render worker never sees it.
- Client cloud (Mode 2) is deliberately not called zero-egress: content reaches the LLM provider you bring, inside your own cloud boundary, under your own contract.
Side by side
| Criterion | Mode 1 — On-prem | Mode 2 — Client cloud | Mode 3 — DocMark cloud |
|---|---|---|---|
| Where it runs | Your own hardware / network | Your own cloud tenant | DocMark's managed Azure |
| Who runs & manages it | You | You | DocMark |
| AI / LLM | None as installed — needs no internet; AI only if an operator enables it | Your own LLM subscription (BYO; we recommend options) | DocMark-managed, included |
| Who authors content | Your people (AI only if enabled) | AI via your LLM, or your people | AI or your people; your copilot via MCP |
| What "zero-egress" means | As installed, DocMark sends nothing off the box — keeping the box off the internet is your network's job | Not called zero-egress; content stays in your tenant | No PII reaches the execution plane |
| Verification gate | local — deck gate is on by default | in your tenant — deck gate is on by default | |
| Brand provenance audit | |||
REST API (/v1) + MCP | You run it locally | You run it in your tenant | Available now — included in every plan |
| Branded decks | Per engagement | Per engagement | Live today |
| Documents · spreadsheets · newsletters · brochures · video | Per engagement | Per engagement | Live today |
| Web one-pager | On the roadmap — not available yet | ||
| Best for | Air-gapped confidential content | AI authoring inside your own cloud boundary | Fastest start with managed AI |
Not sure? Walk the tree
Confidential-content teams that need an air gap start with Mode 1 — On-prem. Teams that want AI but must keep content in their own cloud boundary choose Mode 2 — Client cloud and bring their own LLM. Teams that want the fastest start with managed AI choose Mode 3 — DocMark cloud.