Editions
NEURALVAULT
EDITIONS

Three editions.
One application underneath.

Free and Personal are the same application: moving up unlocks parts that are already installed, so it is a licence file rather than a reinstall and nothing has to be migrated. Business is that same harness installed on your hardware, with the server half a team needs. Here is what each one is, in order, and then the three side by side.

  • Windows
  • v1.0 beta
  • Runs on your hardware
  • Your own API keys
Edition One

NeuralVault Free

The same chat window you already use.
It just keeps what it learns.

You install it, you open a conversation with Cortex, and you point it at your own material. The difference from every other chat window is what happens after you close it: what it worked out is written to a folder of Markdown on your disk, and the next conversation starts from there rather than from nothing. Nothing expires, no account exists, and the paid capabilities are locked rather than missing, so the day you want them it is a licence file and not a reinstall.

What you do with it

Point it at notes, research, client files or a repository, and ask it things that need all of them at once.

Where your material goes

Onto your own disk, as files you can open in any editor. Run the model locally too and none of it leaves the machine.

What it costs to find out

A download. No card, no trial clock, and no email required before you can run it.

Even the free one

How it stands against a plain chat assistant

Where it matches one
  • Chat with a resident assistant, with history and voice input
  • Memory held as plain Markdown you can open and edit
  • Rewind files to an earlier state
  • An audit log of what ran, kept beside the vault
And then it adds
  • The model can live on your hardwareNeuralVault has no AI in it, so you give it one, and you choose where that one runs. Either point it at a cloud provider with your own API key, or install a model locally through Ollama and run it on the machine in front of you. A chat window only ever offers the first, on somebody else’s servers. You can also do both and pick per job, so the cheap work runs locally and the hard work goes out. Keys sit in Windows Credential Manager rather than in a plaintext file.
  • Everything you know, graphed into a lasting local indexEvery note a node and every link between them an edge, laid out in roughly 100 to 200 milliseconds when the vault opens. Search runs lexically and semantically at once, so an exact string and a half-remembered description both find it, and an answer points at the note it came from rather than sounding plausible.
  • Six connectors, on the roadmapShopify, YouTube, Meta Ads, Search Console, GitHub and RSS, reading into one place so the numbers are current when you open them rather than when somebody last built a report. Not yet in the shipping build. Coming soon, on every edition including this one.

Those three are what a free build can claim. The paid editions add nine more. See all twenty capabilities measured side by side.

And what you are missing

Free is still missing the best of NeuralVault.

  • The agent fleet: Cortex splitting a job across agents and sub-agents
  • The board, the workflow builder, Skills and Watch & Learn
  • Anything that starts on its own: the cron scheduler, file triggers, the board runner
  • The approvals queue, Canvas and Calendar
  • The plugin API and lifecycle hooks
  • Sync across your own devices

This is the half that makes it a harness rather than a chat window, and it is the half worth paying for. Locked, not absent: same build, so unlocking is a licence file rather than a second download, and everything you put in the vault while you were on Free is still sitting there afterwards. That half is Edition Two.

Edition Two

NeuralVault Personal

Free gave you the conversation.
Personal gives you the whole application.

Same install, same vault, same conversation. What the £20 turns on is every screen that conversation was talking to: the board agents work themselves, the fleet that splits a job up, the schedule that starts one without you, and the approvals queue that holds anything it wants to send. It is the complete harness, and nothing in it is held back for a higher tier.

It stops waiting to be asked

A cron schedule, a file landing in a folder, or the board itself can start a job. Free answers when you type. Personal is running when you are not.

One brief becomes many jobs

Cortex takes an objective and splits it across agents, and those agents split it again. What comes back is the finished thing, not the first attempt at it.

You approve rather than supervise

Anything outbound waits in a queue with the reasoning attached. You are the last step, not a spectator at every one of them.

In detail

What the twenty pounds actually turns on

  • Agents that split work, and loop over what failedHand an agent an objective and it breaks the job across sub-agents of its own, sees the tools each one is allowed to touch, and iterates over what actually failed rather than stopping at the first attempt. It stops on its own, and the diff comes back for approval rather than as a fait accompli.
  • The whole application, not just a coding toolA board your agents work, a workflow builder over plain Markdown, an approvals queue, a calendar, voice memos and a screen recorder that turns a session into a procedure. All included, none of it an upgrade.
  • The setup you already wrote, read as it isYour CLAUDE.md, AGENTS.md, skills, slash commands and MCP servers are picked up from where they already live. No migration, no second file to maintain, and nothing to rewrite into a proprietary format.
  • Two models marking each otherOne model answers, a second is handed the answer alone and told to find what is wrong with it. Run them from different vendors, or either against something on your own hardware, because a model checking its own work is the weakest check there is.
  • Memory that survives the sessionIt stops re-learning your stack every morning. Corrections stick, and it gets sharper the longer you run it rather than resetting every time you close the window.
Counted, not summarised

Every surface it unlocks

Thirty of them, and every one is a named row in the product's own editions table rather than a paraphrase. Free already carries the vault, the search, the chat, the audit log and local models. This is what sits on top, and it includes MCP and the six connectors still on the roadmap.

Where the work happens
  • Board
  • Workflows builder
  • Skills
  • Canvas
  • Calendar
  • Approvals queue
  • Vault browser
What starts without you
  • cron scheduler
  • file triggers
  • autonomous board runner
  • lifecycle hooks
  • session handoffs
  • sync across your devices
What the agents get
  • 26 sandboxed tools
  • per-agent memory
  • agent-to-agent room
  • agent budgets
  • self-rewrite review
  • rolling chat compaction
What you can see
  • Dashboard
  • Income
  • Activity
  • agent run history
  • run ledger
How you feed it
  • Watch & Learn
  • Voice Memos
  • push-to-talk voice
  • spoken replies
  • Markdown folder import
  • Plugin API
What the money buys

£20 is the software. The AI is separate.

Today: you supply the model. NeuralVault is the harness, so the £20 buys the memory, the agents, the permissions and the record, and nothing else. Give it a brain by connecting a cloud provider on your own API key, or by running a model locally on your own hardware, or both at once with a different model per job. Until you do one of those, nothing answers.

Coming: buy the calls from us. Holding your own provider account is the step most people never finish, so managed routing is meant to become the ordinary way in: pick a model from a list, nothing to connect, from around £50 a month with the software included, and a one-off top-up if you run low. Local models keep working alongside it, so you can still run the cheap work yourself. None of it is built yet. The full ladder is on the Personal page.

And what does not change

No hidden meter on top

  • £20 is the starting price. Bring your own API keys or run a local model and it is the whole bill
  • Memory is still a folder of Markdown on your disk, not a row in somebody else’s database
  • Run the model locally and nothing you type reaches an outside provider at all
  • Install it on the machines you control, with one session running at a time

Still one person and one machine: sync is across your own devices, not across a team. The moment somebody else needs to see the work, that is the next section. The full argument for Personal, and what it costs.

Edition Three

NeuralVault Business

Personal is one person's harness.
Business is that harness, shared.

It includes everything in Personal, unchanged. Every agent, every screen and every capability from the edition above is here too: nothing is taken away and nothing is repackaged to make room. What Business adds on top is the layer that only starts to matter once the work is not yours alone: accounts, permissions scoped to folders, agents scoped to roles, and a record of who did what. That is the part nobody needs on their own machine and nobody can run a company without.

In detail

What changes once other people are in it

  • People, each with their own reachPer-user accounts and folder-scoped permissions on a sync server you host. Finance sees the accounting agent, nobody sees more than their role allows, and adding somebody does not mean handing over everything.
  • Permissions set per action, not per personReading is usually open. Anything that spends money, changes a price or emails a client stops and waits for a named person. An agent working for somebody cannot reach what that person cannot reach.
  • A record you can actually auditEvery action logged with what ran, when, and why, in an append-only log beside the vault, and exportable. When somebody asks what the AI did in March, there is an answer rather than a shrug.
  • Team comms beside the workWire brings channels and email into the same place as the tasks instead of a separate tab. It reads and drafts today; anything drafted waits in the approvals queue rather than sending itself.
Where it runs

One server, in your building, and everyone with their own view of it

  • Installed on hardware you own. If the machines you already have are enough, we say so rather than selling you a server.
  • One vault, one server, and a separate account per person. The reach is per role; the material is not copied around to make that work.
  • Run the model inside the building too and nothing you type reaches an outside provider at all. Use a cloud model on your own key if you would rather.
  • Anything outbound stops in the approvals queue with its reasoning attached, so nothing leaves until a named person sends it.
Modular

Not all of it is switched on by default

Always
  • A sync server, with per-user accounts and folder-scoped permissions
  • Role-scoped agents, so nobody reaches past what their role allows
  • An audit trail across the whole organisation, exportable
Only if you want it
  • Business metrics: revenue, ad performance and channel stats on one timeline
  • Wire, bringing team channels and email alongside the work
  • Deployment onto dedicated hardware inside your building, with the security review that goes with it

That is why there is no single number on the card below: a five-person deployment with none of the optional layer and a hundred-person one with all of it are not the same thing. The consultation is where that gets settled, and it costs nothing. The full argument for Business, or arrange the consultation.

All Three, Side By Side

Start free.
Pay when it has earned it.

Each column contains the one before it. Nothing is removed as you move up and nothing is held back to make a higher tier look better: Personal is the complete harness, and Business is that same harness with the parts a company needs switched on. Which is why the cheapest way to choose is to install the free one and find out.

Start hereChat-first, and yours to keep

NeuralVault Free

£0

No account, no card, and no expiry

For finding out whether an assistant that remembers is actually different from one that forgets. It is the same binary the paid editions run, with fewer switches on, so nothing you learn here is wasted. Like every edition, it has no AI in it: connect a provider on your own key, or run a model on your own hardware. Without one, nothing answers.

A working slice of Personal, not a demo of it.

no cardno expirysame build
  • Chat with CortexThe full conversation surface, pointed at your own material rather than at a stranger’s.
  • A Markdown vault you can searchNotes held as plain files on your disk, searched by keyword and by meaning at the same time.
  • Your choice of modelA local model through Ollama, or Anthropic, OpenAI and others on your own API key. No gate on either.
  • Rewind any run, byte for byteFiles are checkpointed on every write, so a bad run is undone rather than argued with.
  • An audit log of every tool callWhat ran, when, and why, written beside the vault rather than into somebody else’s telemetry.
  • The vault browserOpen, read and edit the whole thing from inside the app, or from any editor you already use.
  • ConnectorsShopify, YouTube, Meta Ads, Search Console, GitHub and RSS, on every edition including this one.
Not in Free
  • The agent fleet, the board, and anything that runs on a schedule without you
  • The approvals queue, the plugin API, and sync across your own devices
Download NeuralVault Free

The installer is not code-signed yet, so Windows shows a blue “Windows protected your PC” panel the first time you run it: choose More info, then Run anyway. The certificate is in progress and this notice goes when it lands.

Best for one personFor your own work, on your own machine

NeuralVault Personal

From £20

a month, excluding VAT

For one person doing their own work: a developer, a researcher, a consultant, anybody with material they would rather not paste into a browser tab. This is the complete harness, and nothing above it is a better version of it. £20 is the starting price, and with your own keys it is the whole bill.

Everything in Free, and then:

your own keysno per-token billevery screen
  • The full agent fleetCortex splits an objective across agents and sub-agents instead of answering it alone.
  • The board, workflows and skillsA kanban board agents work themselves, a workflow builder, reusable skills and Watch & Learn.
  • Work that starts without youA cron scheduler, file triggers and an autonomous board runner, so a job can begin while you are asleep.
  • Dashboard, Canvas and CalendarIncome and activity on one screen, next to a canvas and a calendar the agents can read.
  • An approvals queueAnything outbound waits with its reasoning attached until you clear it. Nothing sends itself.
  • Watch & LearnDo a job once with the recorder on and the agents have the steps for every time after that.
  • Plugin API and lifecycle hooksA local socket any language can talk to, and your own commands run at the start and end of every turn.
  • Sync across your own devicesThe same vault on every machine you own, without a server in the middle.
What Personal actually does

Not on sale yet. £20 buys the application, not the AI. There is no model inside it: run one locally on hardware you own, or connect a provider on your own API key and they bill you for the calls. Buying the calls from us instead is a separate purchase and starts around £50.

SupersetFor running a company, solo to 100+

NeuralVault Business

Let's talk

Priced by size, and by how much of it you switch on

For the point where the output stops being yours alone. Everything Personal does, plus the accounts, folder scoping and record that make it safe for other people to work inside it. Scoped, installed and supported by us rather than downloaded and hoped for.

Everything in Personal, unchanged, and then:

your hardwarescoped per rolea fixed number
  • Business metricsRevenue, ad performance and channel stats pulled onto one timeline that updates itself.
  • A sync serverPer-user accounts and folder-scoped permissions, so each person sees their own work and no more than that.
  • Role-scoped agentsFinance sees the accounting agent. Nobody sees more than their role allows.
  • Team commsWire brings channels and email alongside the work instead of in a separate tab. It reads and drafts today; drafts wait in the approvals queue rather than sending themselves.
  • A security reviewWhat the agents may reach, who may see what, and what leaves the building, checked before anything runs.
  • Deployment and supportThe whole thing built onto dedicated hardware inside your organisation, and kept running.
  • More than one vault on one serverNamed vaults on the same deployment, so departments or clients stay separate without a second machine.
  • Clear the queue from anywhereThe approvals queue is reachable over HTTP, so a pending action can be dealt with from another device rather than only from the desk it started at.
  • Backups on an intervalThe vault snapshotted on a schedule you set, onto storage you control, without anybody remembering to do it.
What Business actually does

Two things get priced: the application, and the deployment around it. You get a fixed number before any work starts.

Before you download

What it needs to run

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

Windows

Windows 10 or 11, 64-bit. There is no macOS build, no Linux build and no ARM build.

The download

About 8 MB, and it installs for your user account only, so Windows never asks you for an administrator password.

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. If your machine has never had a Microsoft Edge WebView2 runtime on it, the installer fetches one during setup, so keep it online for that first run. Windows 11 already includes it.

What the price buys

The subscription is the app. The model is yours to supply.

NeuralVault has no AI inside it, on any edition. What you are paying for is the software: the memory, the agents, the permissions and the record of what ran. Model usage is not part of it. Personal starts at £20 a month for that licence, and before anything answers you, you point the app at a model and that half is billed by whoever runs it.

Run a model yourself

Install one locally through Ollama and it runs on the machine in front of you. No provider is involved, so nothing is billed per message. What it costs you instead is the hardware above.

Or use your own API key

Connect Anthropic, OpenAI or another provider on an account you hold. They bill you directly at their own rates, and the licence stays the whole of what we charge you.

Buying the calls from us is separate

Managed credit, where you pick a model from a list and pay us for the usage, is its own purchase on top of the licence rather than something the licence includes. It is not on sale yet.

That choice is the reason the requirements above matter. On your own API key the model does its thinking somewhere else and your machine only has to run the app. On a local model your machine does both, so the model you pick, rather than the licence you bought, is what decides the real cost of the work.

The step up

What actually changes between them

  • Free → PersonalThe application stops waiting to be asked. Agents, the board, schedules and the approvals queue all switch on, and the same vault carries straight over with nothing to migrate.
  • Personal → BusinessThe work stops being yours alone. Accounts, folder-scoped permissions, role-scoped agents and an organisation-wide record, on a sync server you host.

Free is out now. Personal follows, then Business. Choosing now costs nothing either way: Free and Personal are one download with a licence file on top, and Business is scoped with you before anything is installed. Want it capability by capability? Compare all three against a plain chat assistant.