The same application in the cloud or behind locked doors
The search core is identical in both modes — the $rankFusion hybrid query runs the same way in MongoDB Atlas and in the self-hosted Community edition. Only three swappable adapters differ, and they are chosen by tenant configuration, not by changing code.
Managed operation in the EU
MongoDB Atlas in a European region, embedding and reranking directly in the database, generation via the Claude API. No hardware to buy, deployed in days.
- Automated Embedding — vectors are created inside the database
- Verifiable citations via the Citations API
- No upfront hardware investment, you pay per query
A closed solution on your own hardware
MongoDB Community 8.2 on your own machine, embedding and reranking via Infinity or TEI, generation via vLLM. Neither content nor queries leave your perimeter.
- Suitable for special categories of data and classified information
- Your choice of model — Qwen3, EuroLLM, Gemma and others
- Works fully air-gapped, with no internet connection
Three swappable adapters
| Cloud | On-prem | |
|---|---|---|
Embedding Understands meaning, not just words Turns every paragraph into a numeric fingerprint of its meaning. That is how it finds the right passage even when you phrased things differently from the regulation — ask about “what you pay for a transfer” and it returns the article on transfer fees. | Atlas Automated Embedding (voyage-4) | Infinity / TEI (voyage-4-nano, BGE-M3) |
Rerank A second reading that fixes the order The first pass is fast but rough — it sweeps thousands of paragraphs and picks dozens of candidates. Rerank reads those carefully alongside the question and reorders them so the most precise one ends up at the very top. | $rerank inside the database (rerank-2) | Infinity / TEI (BGE-reranker-v2-m3) |
Generation Writes the answer and shows its sources Composes a plain-language answer from the retrieved passages and attaches the regulation and article behind every claim. When the answer is not in the material, it says so — instead of inventing one. | Claude API (Citations, prompt caching) | vLLM (Qwen3, EuroLLM, Gemma) |
The choice is per tenant — a single installation serves cloud and on-prem customers at the same time.
Two ways to deploy Contineo
Use Contineo as a locked-down company portal, or as embedded intelligent search right inside your website.
Standalone portal on a subdomain
Search and a portal of company content on your subdomain (e.g. search.yourcompany.com). Lockable to signed-in internal users only.
- Sign-in via Microsoft Entra, Google Workspace or other domain-bound SSO
- Access limited to your organisation's internal users
- A portal with an overview of company content in one place
A search icon right inside your website
Add the Contineo icon to your existing site. On click it expands as an overlay layer and replaces the site's regular search.
- One line of code, no changes to your page content
- Overlay in Contineo's design, fits light and dark sites
- Ideal as a replacement for search on a public website
Try it: click the search icon in the sample page.
From content to an automatic website
In upcoming versions Contineo will build a clear website from your sources — an automatic overview of information, projects and rules, always up to date.
Automatic overview
A structured portal generated from your connected sources, with no manual writing.
Projects & information
An overview of projects, documents and updates in one place, always current.
Zero maintenance
A source update is reflected on the website automatically.
Ready to get your content at your fingertips?
We'll show you Contineo on your own sources.
Get in touchoffice@contineo.app