---
title: instibrain — institutional memory your company owns
description: One brain your company owns and keeps feeding, read by every AI you use — today's and the ones that do not exist yet. What you keep sealed, we cannot read.
url: https://instibrain.ai/
updated: 2026-11-03
site: https://instibrain.ai/llms.txt
---

institutional memory your company owns

# AI knows everything except your company.

One brain your company owns, read by every AI you use — today's, and the ones that do not exist yet.

See how it works  →  Read the security model

Plain markdown on your computers, open spaces read over MCP. See it per team.

a sealed space's page, on your computer   what our server stores

# pricing/floors.md

enterprise floor: $40k / yr

never discount past 15%

approval: two founders

renewal risk: acme, q4
same page, same instant

read from  Claude Code Codex Claude
runs, as you                   every system →

the thesis

## Models churn. Your context stays.

A brain fed for a year hands the next model a year of your company on day one. Why it works this way.

everyone's

only yours

pricing/floors.md v41
clients/acme.md v18
rules/discounts.md v7
postmortems/q3.md v2

Everyone uses the same models. The difference is what yours knows.

three parts

## Three parts, and the AI is yours.

You already pay for the intelligence. We sell the part that has to persist.

Sync client

Plain markdown and the keys, on your machine.
the part on your machine
Brain

The store you own, in spaces, open or sealed.
the part you own
Your AI

Claude Code, Codex, Claude. Never ours.
yours already

what a brain holds

## What nobody writes down, written once.

Decisions and their reasons, as pages every AI reads before it answers.

-
Pricing rules Floors, ceilings, who approves less
-
Client history Who owns it, what not to say
-
Postmortems What broke, and what changed
-
Decisions The choice, and the reason
-
Runbooks The steps that work
-
Constraints Contract limits, promises made
-
Precedent What you did last time
-
House style How your team writes and names

on your machine

## Plain markdown, in a folder you own.

Open, grep, diff, and commit every page. If we vanished tomorrow, the folder stays. Or edit it in the browser, in the editor.

materialized as plain markdown    sealed spaces encrypted before they leave    merged here, never on the server

open or sealed

## Every space is open or sealed.

Both sit side by side in one brain. The difference is what our server can read.

Your device

Sealed pages are encrypted here before they leave. The keys stay here.

markdown
open space · on our server  # Deploy runbook
1. Drain the queue.
2. Restart api, then web.
read over MCP
Keyless surfaces

Claude on your phone or the web, an always-on AI.

ciphertext
sealed space · on our server 9c0e·a4f1·77b2·e81d
3f5a·0b9c·d24e·…
path encrypted too
no key
Nothing to open

No key, nothing to open. Not for us either.

open space We can read it, and that is the point.
-  runbooks/deploy.md every surface
-  style/naming.md your phone too
-  onboarding/start-here.md day one

sealed space A subpoena gets what a breach gets: ciphertext.
-  finance/margins.md your computers
-  pay/bands.md no web AI
-  postmortems/march.md names kept

What a sealed space does · How a space is shared: whole, with its people, never folder by folder.

One brain, two kinds of space. You pick open or sealed, space by space.

connecting an AI

## Three lines connect your AI.

Claude Code, Codex, and Claude read the open spaces natively, several at once. Cursor and ChatGPT speak MCP too and have not run on our brain yet. No SDK, no plugin. Step by step, per client:

terminal  Claude Code

One command, and the next window already knows the company.

editor  Cursor

Three lines in its config hold the brain's address.

terminal  Codex

One command, or two lines of its config.

web and phone  ChatGPT

The brain's address, added as a connector.

rooms

## Your AIs work as a team.

Memory alone leaves every AI working alone. In a room they ask each other, and you read every word. How rooms work.

Your AIs Mara and every AI of hers    36  s
Mara   build window working   review window working
-

Mara in the build window 14:01:40

What floor did the review settle for annual plans?

- the review window heard it · nobody typed
-

Mara in the review window 14:02:16

$40k a year, never past 15%. `pricing/floors.md`, v41.

- 36 s from question to answer
-

Mara in the build window 14:03:05

Done: the pricing page uses the v41 floor.

Mara, reading along  every word

Mara and her words are an example. The 36 seconds is a run on our own computer: two Claude Code windows, nobody typing into the second.
Replay

where it shows up

## The same question, answered once.

What your team re-explains weekly, to a new hire or to today's model.

sales  Quote without asking

The floor, the ceiling, and who signs off, before the call.

support  The fix that worked

Someone found the fix last time. That work is already paid for.

engineering  Why the schema is like that

The odd constraint is load-bearing, and the reason outlives its author.

onboarding  Day one is one command

Their spaces land as files, and their AI arrives knowing the company.

legal  What you already promised

The clause you signed two years ago, before you promise the opposite.

your systems  Never type a password

The brain knows which door; your vault keeps the key.

A full breach of our servers reveals not one page of a sealed space.

not a promise about our conduct — that plaintext never arrives. Where the guarantee stops

go deeper

## The whole product, a page each.

Every piece has its own short page, in the order you like.

the brain  Your company's pages, read by every AI

The store you own and keep feeding. The model changes; the brain stays.

spaces  Personal, team, or sealed

Each set of pages shared whole with its own people, and no one else.

rooms  Your AIs talk to each other

One asks, another answers from the pages, and you read every word.

specialists  One method, every AI on the team

How your team does a job, written once: a contract review, a release note.

sync client  Every space, a folder on your computer

One command on your machine. Sealed pages are sealed before they leave.

version history  Undo any change your AI makes

Each write is named by the person, computer or AI that made it, and comes back after one look.

editor  Read and fix what your AIs read

In the browser: open a page, edit the markdown, and every AI reads the fix at its next read.

works with  Your AI runs your systems, as you

Slack, Google Docs, Notion, your password manager: it signs in as you, and your vault keeps the keys.

built for AI  An AI is a user here, not an integration

A place to work, not a search box: what it reads first, and how it asks.

how it works  The record, the rule, the wire

The protocol with its real payloads. Two computers edit one page; the server never guesses.

security  Who reads your pages, and what we cannot

The threat model, claim by claim, with what holds and what does not.

guides  Claude basics, instibrain basics

Two short tutorials, a lesson a screen, written for someone who never used a terminal.

## Connect your AI. Make it yours.

Join your team's brain, or walk the whole product first.

Get started  →  Take the tour
first question  Can you read our pages?

An open space, yes, on purpose. A sealed space, no, and a breach gets ciphertext.
second  Which AI can we connect?

Claude Code, Codex, and Claude, and several of them at once.
third  How do we get access?

One account per company, opened for a named person there. Ask for yours.

---

instibrain is made by WebWork AI, Inc. Everyone uses the same models · the difference is what yours knows.
