OpenMatter

MASKED COMPUTE

Run one model over data nobody has to share.

Train models, score records and compare results with other organisations — without any of you handing over your rows.

Every plan starts you with monthly compute credits. Shared computations are included on every plan.

HOW IT WORKS

Three steps, and your rows never move.

  1. STEP ONE

    Your records stay on your own machine

    Point us at a file, a database or a bucket. It is split into random-looking shares on your computer, before anything goes out.

    • You keep custody of your data the whole time
    • Only shares leave — shares alone rebuild nothing
    • Everyone taking part runs their own copy

    Worth knowing: The shape of the data is public: column names and types are shared so everyone can agree what they are comparing. The rows never are.

  2. STEP TWO

    The answer is computed on the pieces

    A group of machines works on the shares together. None of them ever holds enough to reconstruct anyone's records — and the result comes out anyway.

    • No one sees another participant's records
    • Nothing is pooled in one place
    • Everyone gets the same full result

    Worth knowing: The result is revealed to every participant, so agree in advance that the answer itself is safe to share. Only the built-in analyses can be run.

  3. STEP THREE

    Run the rest of your stack here too

    Databases, dashboards, local models and agent frameworks launch from a catalog, with their passwords and API keys encrypted in your browser before they are stored.

    • Ready-made setups for well-known tools
    • Secrets are encrypted before they leave the page
    • Bring your own model provider keys

    Worth knowing: You supply your own accounts and keys for anything third-party.

ALSO INCLUDED

Launch 40 ready-made setups

Databases, dashboards, local models and agent frameworks — configured, with their passwords and keys encrypted in your browser before they are stored.

  • Ollama
  • PostgreSQL
  • Grafana
  • n8n
  • Qdrant
  • Redis
  • code-server
  • Metabase
  • + 32 more

PRICING

Pick a plan. Move up when you need to.

Shared computations with other parties are included on every plan. Move up as your organisation needs more.

GET STARTED

Which of these is you?

I'm signing up for myself

A personal workspace you can invite people into later. Pick a plan and move up when you need to.

  • Set up in a minute — no forms to fill
  • Monthly compute credits included
  • Add people whenever you want

I'm setting this up for my team

A shared workspace with roles, projects and one budget. You'll be the owner, and can invite colleagues as soon as it exists.

  • Roles and per-project permissions
  • One budget across everyone's work
  • Pick your plan as you set up

QUESTIONS

Before you sign up

Do I have to move my data anywhere?

No. It is split into shares on your own machine and only the shares go out. The column names and types are shared so everyone can agree what is being compared, but the records themselves are not.

What does a plan include?

A workspace, monthly compute credits, the full application catalog, encrypted secrets and the model gateway. Shared computations with other parties are included on every plan.

Do I need a card to start?

Yes. A card is needed to start any plan. Where a plan offers a trial, nothing is charged until it ends, and you can change or cancel the plan from Billing.

Do I need a wallet to sign in?

You sign in with an email address or an existing account, the same as anywhere else, and pay by card. If you'd rather connect a wallet you can, but nothing here requires one.

Can I invite people later?

Yes. Both routes create a workspace you own; you can invite people into it whenever you like and give them roles per project.

What do I need installed?

Nothing for deploying applications — that all happens in the browser. Taking part in a shared computation needs a small agent running on your own machine, which runs on Linux and Apple Silicon Macs today.