Back to home

Your therapist remembers. Your AI should too.

July 16, 2026 · Use cases

Why continuity of personal context across sessions matters — and why it shouldn't require handing your inner life to a chatbot company.

← All posts

A good therapist remembers. Not everything — but the shape of things. The recurring theme. The name you keep coming back to. The thing you said six months ago that suddenly matters now. You walk in, and you don't have to start from zero. The continuity is the work.

This post isn't about replacing that. It's about the gap next to it — the gap where, between sessions, between tools, between the people you talk to, you keep re-explaining yourself. To your AI. To your journal. To the next app you try. The context never accumulates. You start over every time.

That gap is what Eigen Mesh is for.

The thing nobody warns you about with therapy-adjacent AI

People use AI to think. That's not a controversial sentence anymore. You talk to an assistant to untangle a thought, to rehearse a hard conversation, to figure out what you actually feel about something. It's useful. It's also, quietly, a context-destroying habit.

Every new chat is a blank slate. You open a window, you explain the situation, you get something back, you close the window. Next week you open a new one. You explain the situation again. The assistant has no idea that the situation is the same one. You're doing the emotional labor of context-loading every single time.

Worse: the memory features that do exist in those assistants build a profile of you from your chats — a profile you can't see, can't edit, can't take with you, and that's used to train their next model. You're handing your interior life to a company in exchange for a feature that doesn't even work across tools.

Eigen Mesh is built on the opposite assumption. You say it, you meant it, it's saved. The memory is yours. It lives in one place you control. Your assistants — whichever ones you use — come to it.

What continuity actually looks like

Here's a concrete version. You're working through something with a therapist. Between sessions, you use an AI to think out loud. You capture the things that feel important in Eigen Mesh — not transcripts, not diaries, just the sentences you want to keep.

"The thing about my dad isn't the argument. It's that he didn't call after."

"I think I'm not actually angry about the job. I think I'm tired."

"The pattern I keep noticing: I shrink when someone else is louder."

A few of these a week. Maybe one a day. Over a few months, you have a shape — not a diagnosis, not a profile, just the things you've chosen to keep. When you sit down with your AI to think, you can ask: "What have I been working on?" and the AI can pull from your memory. You don't re-explain. You continue.

This is the part that matters: it's not the AI remembering you. It's you remembering yourself, with the AI holding the string.

A note on what this is not

This is not medical advice. It's not a diagnosis tool. It's not a replacement for therapy, and it's not a journal your therapist can read. It's a memory of the things you chose to say, in your own words, available to the tools you chose to use.

If you're in crisis, talk to a person — a licensed one. Eigen Mesh is for the in-between. The thinking-out-loud. The "I want to remember what I noticed about myself this month."

Why deliberate capture matters here more than anywhere

The temptation, with anything therapy-adjacent, is to let the AI infer. Read your journals, your messages, your emails, build a model of your emotional state. That's the easy path. It's also the path that produces the worst dataset: inferred, uneditable, and owned by someone else.

Eigen Mesh goes the other way. You say the thing. You hit save. It's there. Nothing else is. The dataset is small — a few sentences a week — but it's exactly what you meant, not what an algorithm thinks you meant. When you come back to it in six months, you'll recognize the voice. It'll be yours.

That's the whole bet. A memory that's smaller but truer beats a memory that's vast and inferred.

The portability part

The reason this works across sessions and across tools is MCP. Eigen Mesh speaks it. So when you're in Claude, working through something, Claude can read your memory. When you're in Cursor, drafting an email you've been avoiding, Cursor can pull the context. When you switch to a different assistant next year because something better came along — your memory goes with you. You're not locked in. The memory is.

This is the difference between a memory feature and a memory you own. The feature dies with the product. The memory outlives the product.

What it costs

100 free credits — about 100 to 150 thoughts. That's months of weekly reflection, easily. After that, €3.99 a month. Self-host it for free if you'd rather run it on your own hardware (Apache 2.0, github.com/hikaru90/eigen). Hosted version runs on EU servers. We don't train on your data. The things you said about yourself stay the things you said about yourself.

If you're tired of starting over

If you've ever closed a chat window and thought, "I just explained all of this last week" — you have. You will again. Unless the memory holds it.

Get early access to Eigen Mesh. Say one true sentence. See if it comes back.

Eigen Mesh

Open source memory infrastructure — Apache 2.0, self-hostable, and less than a coffee a month on managed hosting. Your context stays yours.

Navigation

© 2026 Eigen Mesh. All rights reserved.