Lightweight
An 8.5 MB executable that runs on a laptop or a server, with no vector database, no Python and no containers.
RAGNIX · DOCUMENT AI
Your assistant answers with the manual in hand. And the manual never leaves your network.
Cite or abstain. The whole procedure or nothing. Never from memory.
REQUEST A DEMO
In pilot
IN SHORT
What is RAGNIX?
RAGNIX is a document retrieval engine for organisations whose documents can’t leave their network. It indexes your manuals and serves them to the AI assistant you already use, over MCP: complete procedures, cited to the document, revision and section. If it finds no support, it doesn’t make things up: it says so.
THE PROBLEM
What problem does it solve?
- An assistant always answers. When it doesn’t know, it makes things up with the same confidence it has when it’s right.
- Half a procedure is worse than none. Steps 3 to 7 of a 12-step procedure are worse than no answer.
- The knowledge lives in one person. Your technicians call «the one who knows», and the one who knows is retiring.
- Your manuals can’t go to the cloud, and in the field there isn’t always coverage.
HOW IT WORKS
How does it work?
Controlled ingestion
Markdown, Word and HTML. Nothing goes in without passing a quality check and your approval, and every change is logged.
Local, versioned index
Search that combines exact words and meaning. Each index version is immutable: if something goes wrong, you roll back to the previous one.
Answers over MCP
Your assistant searches, receives the whole section with its citation, and browses the document’s table of contents. It works with small, local language models.
GUARANTEES
What does it guarantee?
- It never hands over half a procedure.
- Every answer carries a verifiable citation: document, revision and section.
- It never degrades silently: if something fails, it says so.
- Nothing leaves the network.
WHAT SETS IT APART
What makes it different?
No per-query cost
And no dependence on a cloud AI provider.
Access control per fragment
Isolated collections, and restricted content is hidden by default.
No telemetry
It sends us nothing.
USE CASES
What is it used for?
| Sector | What it solves |
|---|---|
| Industry and utilities | The technician looks up the equipment manual in the field, with the whole procedure and its citation. |
| Operations and maintenance | What the expert knows, available to the whole shift. |
| Defence and security | Assistants that consult documentation that can’t leave an air-gapped network. |
| Any organisation | An assistant over its internal documentation without uploading it to the cloud. |
COST
How much does it cost compared with a long-context model?
Estimate, not a measurement
- With a 200,000-token manual, a two-question session costs about 24 times less with RAGNIX than sending the whole manual to a long-context model in the cloud (public prices, September 2026). Output tokens are not included. The experiment with a real manual is under way.
What we don't promise
Better answers than putting the whole manual into a giant cloud model. What we promise is that, with several manuals, offline and with a small model, the answer still arrives with its citation.
DEPLOYMENT AND PRICE
Where is it deployed and how much does it cost?
- On your infrastructure: laptop, server or air-gapped network.
- Standalone or integrated into your projects and our solutions.
- Price: by agreement.
STATUS
What stage is it at?
In pilot with a first customer. Today it indexes Markdown, Word and HTML and extracts the figures of each revision. On the way: PDF, serving figures over MCP, encryption and authentication of the MCP channel, and access with your company’s SSO.
FAQ
What people usually ask us about RAGNIX
What is verifiable RAG?
A system that searches your documents for the information the assistant needs and hands it over with its citation, so anyone can check where the answer comes from.
Do I need a specific AI model?
No. RAGNIX connects over MCP to whichever assistant you already use, and it is designed to work with small, local models.
Does it work without internet?
Yes. Everything runs on your network: indexing, search and serving.
Which formats does it support?
Markdown, Word and HTML. PDF is on the way.
Who decides which documents go in?
You do. Nothing goes in without approval and every change is logged. If an index version doesn’t convince you, you roll back to the previous one.
Why not upload the manuals to a cloud assistant?
Because many can’t leave your network, because in the field there isn’t always coverage and because paying to re-read the whole manual on every question gets expensive.
REQUEST A DEMO
Do your technicians call «the one who knows»?
We’ll show you RAGNIX with one of your manuals.
REQUEST A DEMO