Mnemosyne OS
New The documentation is live: every engine, step by step. docs.mnemosyne-os.io →

Dossier 03 · Public sector & institutions

Your archives, searchable, in your jurisdiction

Mnemosyne OS keeps your administration’s document memory on your own hardware, in your jurisdiction. No foreign cloud is imposed. From a town hall to a ministry, this is the default setting.

Do you have data-residency obligations? Let us talk about on-premise deployment and auditability.

Windows, macOS and Linux.

Local, hybrid or your server

Your data stays in your jurisdiction

100% local

Everything on the machine

The model runs inside the application (Qwen, Gemma, Phi in GGUF). Unplug the network cable: search and answers keep working.

Hybrid

Search local, answer in the cloud

Search runs on the machine. A cloud model writes the answer, with your own key or your Vertex project. The question and the retrieved passages go to that provider.

Your server

Your administration’s model

Your IT already runs a model? Connect it: any server that speaks the OpenAI-compatible API, at an address you set in the settings. The question and the passages go to that machine.

Compatible models

  • Vertex AI
  • Gemini
  • OpenAI
  • Anthropic
  • Mistral
  • DeepSeek
  • MiniMax
  • Qwen (local)
  • Your server

An administration can work fully offline on its sensitive archives, and in hybrid mode on public documents. Download Mnemosyne OS to try it on one computer. For a deployment on your servers, talk to us.

Download Mnemosyne OS → Talk to us →

Working across departments

Share vaults on the administration’s network Beta

Every computer in the department runs Mnemosyne OS. On the same network, they connect directly, with no server in the middle.

You lend a colleague a vault to read. Or you share a vault with the team: each person writes to it and keeps their own copy. The computers can also send each other messages.

A deletion never spreads. If two versions of a file diverge, both are kept. Between two sites, the application sets up Tailscale and guides you.

The network stays off until someone turns it on, computer by computer.

Fig. · Four computers in a department, one shared vault

For which body

  • Town, department

    Small body

    A file server, twenty years of records, and nobody to tidy them. Mnemosyne OS installs on a workstation.

  • Agency, region

    Mid-sized body

    Several directorates, several document collections that never meet. One vault per domain.

  • State administration

    Large administration

    On your machines, in your jurisdiction. Deployment, licences and bespoke cartridges on quote.

  • All three

    The same software

    The install and the licence change. The memory does not.

Inside the administration

When an officer changes jobs, their files stay searchable

01

Your records

Twenty years of minutes, reports, public contracts and exports. You point at the folder, Mnemosyne OS reads it in place. Your files do not move and are never modified.

02

Your question

You ask your question in your own language. The answer cites the source document. Two documents on the same subject are linked, even when they use different words.

03

When an officer leaves

An officer changes jobs, and their successor cannot find the files. What the administration wrote stays searchable after its author has left.

04

The links

When the machine is idle, the engine links documents together. A map shows these links. A decision from 2019 often comes back this way, from a forgotten report.

Questions

What we are asked before a file is opened

Where does the data live, and what leaves the machine?
The memory lives on your machines. The engine reads your folders in place, the index is local, and vault content can be encrypted with AES-256 at rest. What leaves depends on the model, and you choose it. With a local model, your documents and your questions stay on the machine. With your own server, the question and the excerpts stay inside your network. With a cloud model, they go to that provider, under your key. Every call to a remote model is written to a local journal, with its tokens and its duration.
Can it run entirely offline?
Yes for the memory: search, vectorization and a local model run with no network. No account is needed to use the software. A connection is only used to download models, install updates and check a licence.
Who can read what?
The memory belongs to a workstation. It is protected by the machine session and by vault encryption. A vault can be marked MAXIMUM: in the chat, its content then only goes to a cloud model if you confirm it, every single time.
How is it bought?
By licence. For a public body, volume licences, on-premise deployment and a bespoke cartridge are quoted. We answer the documents of a tender.
What if we stop?
Your documents stay in place, in their original format: they were never moved or modified. If you stop, you lose the index, not your documents. Nothing is stored at our end, so there is nothing to retrieve.

We have a security dossier that says where each piece of data lives, and an SBOM. Since 12 September 2026, we are registered on ENISA’s reporting platform, with a written procedure. And we host none of your documents: your data is not in our audit scope.

Two features for public service

Find an exact reference, check every sentence

An exact reference

A resolution number, a contract reference, the name of a hamlet. Search combines meaning and exact words. A reference is found as written, even inside a document about something else.

A verified answer

With verification turned on, a second pass compares each sentence of the answer with the cited documents. The officer sees the sentences no source supports before copying them into a letter.

Both features are in the application. To connect them to your records and your access rules, we build the deployment with you.

Contact us →
← Back to the audience router