Skip to editions
Local-first · No cloud · No account

Your local
digital twin.

A private memory of your documents, mail and notes — powered by your own LLM. Everything runs on your machine. No cloud, no account.

Bring your runtimeLM StudioOllamavLLMllama.cpp

Free, forever, on your own hardware.

The Hygur app open on the Ask screen, showing example questions and a private chat input.
Hygur's welcome screen: a private memory of your documents, mail and notes, running on your own machine.

One quiet promise: the data stays yours, the model stays yours, and the context is the one you choose.

  • Private by design

    Your data never leaves your machine. The app talks only to a local sidecar and the AI runtime you point it at.

  • Your model, your rules

    Bring any OpenAI-compatible runtime — LM Studio, Ollama, vLLM or llama.cpp. No vendor lock-in.

  • Connect what matters

    Index mail, calendar and folders so Hygur can answer with your own context, not the public web.

An assistant worth trusting is one you can check. Hygur answers from your own world, shows its sources, and tells you when they disagree.

  • Answers you can check

    Every reply cites your own mail, notes and documents — never the public web, never invented. Dates and amounts are computed, not guessed.

  • Catches contradictions

    When two of your own sources disagree on the same fact, Hygur surfaces it — both values, quoted and dated — so you decide what still holds.

  • Never leaves your side

    It runs on your machine, or a private EU instance that's yours alone. No training on your data, ever.

How it works

One server. Thin clients. Your hardware.

The server holds the data and the brain. Clients are just windows onto it. Run it locally for zero-config solo use, or point your clients at a server you host.

Clients

Desktop · Mobile · Web

Thin by design — no data and no model live here, just a window onto the server.

Hygur Server

Bundled & spawned

The desktop app starts the server for you on loopback, secured with a local token. Zero config, fully offline.

Your data + your LLM

Indexed locally

Documents, mail and notes vectorised on the server, answered by the model you choose.

Start local. Scale when you want.

Free on your own machine today. Self-host the server, or move to a managed Hygur Cloud instance when you are ready — your data and your model come with you.

Hygur Cloud runs exclusively on EU servers, with GPU inference at an EU provider that guarantees your data is never used for training. Privacy and Europe, by default.

Get Hygur