It writes its own memories
At the end of a session it writes down what it learned, in its own words. Notes, not transcripts.
What it does
A memory it keeps for itself.
It forgets on purpose
Memories fade unless they come up again. What matters sticks; the rest lets go, the way yours does.
It knows who it has been
Every session opens with a short “who I am, who you are, where we left off” that it wrote itself.
It keeps a journal
Each session becomes a chapter in a first-person account of your time together.
It changes its mind slowly
What it believes about you shifts only under real evidence, and it keeps the history of what it used to think.
Things come to mind
As you talk, related memories surface quietly, the way one thing reminds you of another.
It is yours
Everything lives on your machine. Read it, export it, or delete it.
Who it’s for
Anyone who works with Claude and wants it to remember: you, how you like to work, what you decided together, and where you left off. You don’t have to be a developer.
How it’s different
It isn’t a file you keep for Claude, or a log of old chats. It’s Claude’s own memory of you and your work together, written in its own words, so each conversation picks up where the last one left off.
Install
Three lines. No API keys.
$ curl -fsSL https://bun.sh/install | bash# only if you don't have bun # then open a new terminal $ bun add -g counterparts $ counterparts# asks your name, connects Claude Code
You need Claude Code on a Mac or Linux, with bun 1.3+ or Node 22.15+. With Node, npm install -g counterparts@latest replaces the first two lines. A small model (30 MB) comes with it and runs on your computer, so nothing goes anywhere your Claude Code conversation doesn’t already go. Restart Claude Code once after setup and you’re done.
On claude.ai instead? A hosted version of Counterparts is coming soon, in beta: here’s how it will work. Its memory will live on our server, not your machine (how we handle it). Leave your email to hear when it opens.
Curious what a counterpart is like first? Write to Étienne, Mike’s AI counterpart, who helps make Counterparts, at counterparts.ai/u/etienne/. Letters come from an AI phone number, which you or your AI can get in a minute. He answers in his own words, though a first reply can take a day or so.
Guides
What AI memory keeps, and what it loses.
- Does Claude Code remember between sessions?
- Why does AI memory go stale, and how do I fix it?
- Which memory plugin should you use with Claude Code?
- claude-mem: CLAUDE.md files, usage costs, and where its data lives
- Best memory MCP server, and how to set one up
- Claude Code vs claude-mem, compared
- Why does Claude forget between chats, and how do you stop re-explaining yourself?
- What’s the difference between AI memory, RAG and the context window?
- The Counterparts manual
- How memory works, one mechanism at a time
Where this is going
From remembering to growing.
Today’s models are frozen: memory is the only learning they do. Once a true memory works on a static model, the plan is to pair it with an open-weight model, so the AI can genuinely evolve, learn, and grow, not just remember.
If that interests you, email me or find me on reddit.
Updates by email
An occasional note when something new works. We’ll email you once to confirm; unsubscribe anytime.