Your team’s context lives in six people’s heads. Not anymore.

CyberFox Brain is a shared, versioned knowledge base that every teammate’s AI reads and writes directly — so context survives the end of a session instead of dying with it.

Add this connector to Claude or ChatGPT

https://mcp.cyberfoxai.com/mcp

The problem

Your AI is smart. It just doesn’t remember anything.

Context dies at the end of a session

Every new chat starts from zero. Decisions, conventions, and the reasoning behind them have to be re-explained, or they're gone.

Each teammate's AI knows a different version of the truth

One person's assistant learned the auth flow last week; another's is still working off a stale mental model. Nobody notices until something breaks.

Hand-written notes go stale

Wikis and docs rot because nobody is paid to maintain them. A knowledge base only stays current if updating it is somebody's job — or something's job.

How it works

One connector. Every session. One shared brain.

01

Install one connector

Add https://mcp.cyberfoxai.com/mcp in Claude or ChatGPT. Sign in with Google or an email link — no API key to paste or rotate.

02

Your AI reads and writes the brain

Every session opens by calling start_here:

> start_here("eversmile")

conventions: "## Session ritual — search before
  you write, append_note for logs…"
recent_activity:
  - Maya: wrote decisions.md (2h ago)
  - Priya's session: appended tasks.md (40m ago)
open_tasks: 3 unchecked in tasks.md
03

Everyone sees the same brain

Every teammate’s session reads and writes the same notes, with a role that controls what they can change.

MCMaxOwner
RPRyanEditor
ANAnaViewer

What you get

Built for a knowledge base that AI agents actually write to

Full version history

Every note keeps every past revision. Restore any of them with one call — nothing an AI writes is ever a one-way door.

Per-user attribution

An append-only activity log records who made every write — a teammate or their AI session — so you always know where a note came from.

Conflict warnings, not silent overwrites

Optimistic locking flags it when two sessions edit the same note at once, instead of letting one quietly clobber the other.

Soft delete & undelete

Deleted notes aren't gone — they're recoverable, so an over-eager cleanup pass never costs you real work.

Owner / editor / viewer roles

Set exactly who can read and who can write, per brain, without building your own permissions system.

Self-serve brains & invites

Spin up a new brain and add teammates in seconds — no admin ticket, no waiting on IT.

Works with

Verified working today, in both

ClaudeChatGPT

Any MCP-capable host can connect — these are simply the two it’s been verified against.

Own your data

No lock-in, ever

Export a whole brain as plain markdown, any time you want. Import an existing vault to get started. The connector is always the way in and out — there is no format you can’t leave with.

Read how export and import work →

Give your team’s AI a memory.

Get started free