Your mind's digital twin
One memory for all your AI assistants.
Coredence reads what you have already written (notes, chats, code) and turns it into memories about you, each tied to the traces it grew from. The mind lives on your computer. Claude Code, Cursor and VS Code read it directly; ChatGPT receives it as text you paste.
macOS 14 or later · Apple silicon · Notarised by Apple
Download for Windows 10 or laterOne memory
A memory, and the traces it grew from
The engine reads what you have already written and keeps what stays true. Every memory carries the moments it formed from, so you can disagree with it about something specific.
This is a real memory from the developer's own mind, measured on 14 October 2026. It grew from more than 200 messages, and it describes a person rather than a project.
“Does not nod along to an explanation that has not landed; asks until it is clear”
Grown from
- “I don’t get what you mean”
- “I don’t get where I’m supposed to add it”
- “still don’t get what I’m meant to do”
- Grown, not told
- Memories form from what you write, not from you describing yourself.
- Every memory has its traces
- Open any memory and read the messages it grew from.
- Lives on your machine
- One SQLite file in your home directory. No account.
How the mind is layered
Five kinds of memory, born from events
The mind lives through events first: every message, note and chat it reads leaves a trace. Memories are born from those traces and kept in five kinds. A preference and a value are not the same thing, and a passing mood is neither — mixing them is how an assistant ends up believing you dislike a library you complained about once, on a bad afternoon, two years ago.
Events
What happened: every message, note and chat the mind has read. Each memory is born from them and opens back onto them.
24,000 events → 604 memories (the developer's own mind, 14 October 2026)
Values
What you will not trade away, however inconvenient.
“Would rather cut the scope than lower the standard”
Patterns
What you do again and again, whether or not you would claim it.
“Finishes the main path before touching the polish”
Voice
How you say things — length, directness, what you refuse to pad.
“Says the uncomfortable version first and softens it after”
Preferences
What you reach for first when nothing forces your hand.
“Would rather push a familiar tool to its limit than switch to a new one”
Current state
What is true this month and will not be true next year.
“Is deep in a launch and guarding the scope hard”
How it works
Download it, let it read, watch the mind form
- 1
You point it at your writing
Notes, a journal, chats with AI assistants, exported messages, and your code and commits if you want them. The corrections you made when something was wrong, too. It asks before it connects anything, and all of it is read on your own machine.
- 2
It consolidates memories, and holds back what it cannot support
Something seen once usually stays a faint trace until it happens again. When something contradicts a memory, it becomes a question for you, not a silent decision. The cycle runs every hour, the way a mind consolidates in its sleep.
- 3
You look at the mind
The whole mind as one neural map: cortices, synapses, memories. Every memory opens onto its traces, and you can correct it, merge it, or say it is not you.
Why you can trust it
What it refuses to do
The engine would rather remember less than remember something it cannot defend. Most of what follows is a decision to stay quiet.
It will not publish what it has seen once.
One trace is a coincidence, so it waits as a draft. The exception is what you state about yourself directly, in a quick note or in answer to one of its questions.
It will not overwrite a memory when something contradicts it.
A contradiction becomes a question you answer. If the memory is confirmed three more times after that, the question closes as an exception and the reason is recorded.
It will not form a memory it cannot show you the traces of.
Every memory keeps its traces or it does not exist.
It will not guess to fill a gap.
A sparse mind is honest. A padded one is not.
It will not upload your mind.
The database never leaves your machine. While consolidating, the engine sends the model you chose the passage being read and your list of memories. Run the model locally and not even that leaves. The privacy page lists every request.
Where it lives
One file, on your machine
The mind is a SQLite database in your home directory. There is no account and nothing to sign in to. You can copy it, back it up, open it with any SQLite browser, or delete it.
The app that draws the map opens that file read-only. It cannot change what it shows you: the operating system refuses, whatever the app wanted.
You never have to take that on faith. While the window is running, ask the operating system what connections it holds: the answer is none.
On a Mac, the first command should print nothing, because the window holds no network connections. The second lists the tables of your mind, in a tool nobody here wrote. A tested Windows equivalent is not published yet.
Where this is going
What ships today, and what does not
The whole point of the product is that it only says what it can support. The same applies to the page you are reading.
01
Shipping now
The engine on macOS and Windows, a window of its own on both, and the connection to the assistants you already use: read your writing, consolidate it, look at it, hand it over.
02
Being built
A view of how the mind changes over months rather than one snapshot.
03
Waiting to open
The plan where we run the model is built into the engine; the sale is not open yet. The pricing page says what it will cost. Today you consolidate with the Claude Code you already have, your own key, or a model on your own machine.
Get it
One download, no terminal
The engine that reads your writing ships inside the app. Nothing to clone, no Python to install, no command to run — open it and it asks what it may read.
macOS 14 or later · Apple silicon · Notarised by Apple
Download for WindowsWindows 10 or later · 64-bit
Open it
It asks which sources it may read, and connects nothing until you agree.
Leave it a couple of minutes
The first pass reads a few hundred events. After that it keeps up on its own whether or not the window is open: reading every hour, consolidating once enough has gathered.
Look
The whole mind as one neural map, every memory opening onto the traces behind it.
Windows works the same way: one download and a double-click. The installer is not code-signed, so Windows will warn about an unknown publisher; the download page explains what to do and how to check the file. A mind is one file and can move between your machines.
Questions
Before you install anything
- Does anything leave my machine?
- Two things. While consolidating, the passage being read and your list of memories go to the model you chose; at most once a week that model also receives some of your memories in order to write questions for you. With a model on your own machine, none of this leaves. Separately, once a day the engine asks github.com for the latest version, and you can turn that off. There is no account, and on the free path no server of ours is involved.
- What exactly does it read?
- The sources you connect during setup, and only those — it asks first and connects nothing until you agree. While the window runs you can check that it holds no network connections, with one command on the privacy page.
- Can I disagree with a memory?
- Yes, and that is the point of keeping the traces. Open a memory, read what formed it, and correct it, merge it, or say it is not you.
- Does it work with my AI assistant?
- Yes. Settings → Assistants lists the clients it finds on the machine (Claude, Claude Code, Cursor, Windsurf, VS Code, JetBrains IDEs, ChatGPT/Codex, Gemini CLI) and connects any of them with one button. Claude Code loads the mind at the start of every session, Codex and Gemini can take it as an instruction file, and every client can query it over MCP. The connection is read-only. The ChatGPT chat app cannot connect to a local server, so the same tab copies your strongest memories for you to paste.
- Is it macOS only?
- No. Windows has the engine and a window of its own, and a mind moves between the two machines as one file. The two windows are functionally equal: the neural map, the layers, search and traces are in both.
- What does it cost?
- Using what exists today is free: you consolidate with Claude Code, your own key or a model on your own machine, and we are not in between. The only paid thing is the plan where we run the model for you when you have no account of your own — an addition rather than a wall around what you already have. That sale is not open yet; the pricing page carries the numbers.
- Is it open source?
- No. The source is not published. The releases, the usage documentation and the issue tracker are public, and the window's privacy claims can be checked on your own machine without the source.
- Who makes it?
- Ahmet Karaca, at Akaya Labs, a one-person studio based in Türkiye.