Download NeuralVault

NEURALVAULT
Download

An agent harness that runs on your own machine, keeps what it learns as Markdown you can open, and answers on a model you chose.

BetaPublic early access. Expect bugs. Keep a backup of your vault.

Latest release: NeuralVault 2.0 Beta, 15 October 2026. Expect bugs. Keep a backup of your vault. See what changed.

One installer for every plan. Your account decides what you get: sign in and NeuralVault runs as Free, Personal or Business.

The choice below is about what happens after you install, not about which file to fetch.

your vault
  • clients/acme/brief.md
  • meetings/2026-09-08-standup.md
  • research/pricing-notes.md
cortex wants to write
append to clients/acme/brief.md
  ## Next steps
+ Confirm scope with Acme before Friday
+ Send the revised estimate, £4,800
model llama3.1 · on this machinenetwork idlecheckpoint taken, rewind available
Nothing is written until you approve it, and every approved write is snapshotted first so it can be put back. Running a local model, as shown, nothing you type leaves the machine. An illustration of the approval flow rather than a screenshot.
Which edition

Same file. Three ways it ends up.

What you download is identical. What differs is the account you sign into afterwards, and that is what each of these pages is about.

Still deciding? Compare the editions, where every capability is marked with whether it works in the next release, is partly built, or is granted by your plan and not written yet. Or read what NeuralVault actually does first.

Beta, public early access

What to expect, and how to tell us.

What to expect from the Beta

  • It will have bugs. That is what a Beta is for.
  • Keep a backup of your vault, and keep copying it.
  • NeuralVault checks for updates and installs them. You can also download the newest installer from this page and run it over your copy; your vault is kept.
  • Windows only. macOS and Linux are experimental, and there is no macOS or Linux installer yet.
  • During the Beta everyone has Personal. In the next few weeks, when the Stable version is released, Personal becomes a monthly subscription. We will tell you before it changes.
  • Personal is all your AI bills in one place: we handle the billing, the routing and the backups, and routing is only available through Personal.
  • NeuralVault routing (390+ models through one account) runs on a prepaid usage balance, and the balance opens soon. Using your own API keys or a local model is free, and always will be.
  • The changelog lists what changed.

How to give feedback

  • Use the bug report inside the application. It takes a minute.
  • This guide shows exactly what a bug report sends and what it hides.
  • For anything else, write to us.

Known issues

  • The licence check can lock the application if it cannot reach our service for seven days. Nothing is deleted, and it opens again when it can check in. What to do.
  • The nv command-line tool is not yet signed, so Windows may warn when you run it.
Before you download

What it needs to run

Free, Personal and Business are the same application with a subscription between them, so the machine needs the same things whichever one you end up on.

Windows

Windows 10 or 11, 64-bit. Windows is the main supported platform. There is no ARM build, and macOS and Linux are experimental (see below).

The install

It installs for your user account only, so Windows never asks you for an administrator password. About 115 MB to download.

On your own API key

The light way to run it. The model does its thinking on your provider’s hardware, so your machine only has to run the app.

On a local model

Much heavier, and worth knowing before you start. Ollama holds the whole model in memory for as long as it stays loaded.

The one local-model number worth having in advance: the model our getting-started guide suggests, qwen3.5:9b, is a 6.6 GB download, and a smaller model costs you proportionally less. We do not publish a minimum memory figure, because with a cloud provider the app is the only thing running on your machine and with a local model it is the model that decides, not us. The app carries its own browser engine, so setup fetches nothing else. The first run needs a connection only to sign in.