AIxVuea private brain for your team and its machines

Every project in its own room.
Your machines do the work.
Your Claude gets one key to one door.

AIxVue is a server brain for a small team: it remembers what each project knows, keeps those memories apart by structure rather than by promise, runs jobs on the computers you already own, and shows the day on a screen on the wall. You can let it drive your Claude Code sessions, or keep your Claude and give it a key.

Open the app See the two ways in
How AIxVue sits between your machines and your projects On the left, your machines: a PC running Claude Code, a lab GPU box, a phone. In the middle, the brain: three rooms, Wavesim Pro, Lab, and a private room, each with a door, and a ledger beneath. Two arrows: A, the brain drives the Claude on your PC and asks you on your phone before every tool. B, your own Claude holds a key and enters one room through the MCP door. Your machines Your PC Claude Code · the claude kit Lab GPU box a kit per tool · Python, solvers Claude asks: run npm test? Your phone The brain Wavesim Pro specs · backlog · decisions three people, one memory Lab experiments runs · results Your own room everything you say, from any device, yours alone door Rights are rows, not promises who · which room · read · write · since · revoked — and a journal of everything that happened A the brain drives your Claude it asks you first B your Claude holds one key MCP · one room · read or write · expires What comes out A plan for today next three items, what is blocked, work done twice Runs on your boxes priced first, approved by you, results kept with provenance A screen on the wall the day, the lists, the house, paired with a code Nothing leaves a room unless a person publishes it into another. The transcript of a Claude session never leaves the machine it runs on.
One brain, rooms with doors, and two ways through them. Path A: you trust the brain and let it drive the Claude on your PC; it asks you on your phone before every tool. Path B: you keep your Claude and hand it one key that opens one room, read or write, until it expires.

Two ways in

They reach the same rooms, the same rights and the same journal. The difference is who holds the initiative, and where the question "may this run?" is asked.

A. Let the brain drive

For a team that wants one screen for all its Claude sessions, with the record kept.

  1. Pair your computer. In the app, open Remotes and run the one line it shows on your machine. It prints a code; name the machine from your phone and choose which room it stands in.
  2. Install the Claude Code kit on that machine from its Kits tab. It needs Claude Code signed in as you, and a Python 3.
  3. Open Claudes. Scan the machine: every session appears as a card, by folder, branch and title. Watch one live, or send it a prompt. Before every tool it uses, it asks you, in the app or on your phone.
  4. Share what matters. Move a session's card into the team's room, or publish its outcome as one fact the team reads.

What leaves your machine: the card and one short line per turn. Never the transcript.

B. Keep your Claude, give it a key

For anyone who would rather not hand control to a server, and still wants the shared memory.

  1. Make a token in the app, under Claudes, Connect your own Claude. Pick one room, read or read and write, and how many days it lives.
  2. Add the door to Claude Code with the command the app shows:
claude mcp add --transport http aixvue https://brain.aixvue.com/mcp --header "Authorization: Bearer axv_…"
  1. Ask it where it is. Your Claude sees that room's memory index, reads facts in full, ticks the backlog, and publishes outcomes. It cannot stand anywhere else: the token names the room.
  2. Revoke any time. A token lists its last use; leaving the room revokes it by itself.

What the key cannot do: grant a right, make another key, or reach a second room.

Screens on the wall

A tablet in the kitchen, a TV in the lab, a sign by the door. Each is paired once with a code and shows only what its room holds.

  • The day. Today's and tomorrow's events with who is at them, what is still open on the shared lists, today's notes. Big type, read from the sofa, nothing to press.
  • The house. Every device as a tile: a tap switches, a sensor shows its reading. The screen reaches the house through a machine standing in it, never through a cloud you do not own.
  • Pair it by scanning. Open the app on the screen, scan the square it shows with your phone, name it, pick its room. A screen's key opens that screen's room and nothing else, and retiring it takes one press.

A member's own room never reaches a shared screen. What is on the wall is what was published to the wall's room.

Scan to open the app on your phone: brain.aixvue.com. Sign in, then pair a screen from the People page.

What holds

These are properties of the code, each with a test that fails when it stops being true.

A room is the boundary

A person without a right in a room never learns it exists: not its name, not its count, not a hint in an answer. Every route and every tool names the right it needs, and a sweep calls every one of them as a stranger with a marker planted inside.

Your data stays yours

Files live in the store a room chose: your Azure or S3 account, your OneDrive, or one you rent. Job results go from your machine straight into that store. The server holds people, rights and the journal, and never a credential of yours.

Nothing acts unasked

Anything that spends or deletes is approved by a person first, priced before it runs. The brain never grants a right, never edits a prompt, never changes code. It gets better only through code, tests, commits and your approval.