You know the ritual. You open a new AI tool. It knows nothing. You explain, again, that you're a designer who works mostly with European nonprofits, that you prefer terse copy, that your client's name is spelled with an ß, that the project you're on has three stakeholders and one of them hates the color orange. You explain. It acknowledges. You close the tool. Next week you open a different one. You explain again.
This is the part of the AI era nobody talks about: the constant re-loading of context. Every app is a stranger. Every chat is a blank slate. The "memory" features in most assistants don't fix this — they build a profile inside that one app, and the profile dies when you leave.
Eigen Mesh exists because memory shouldn't be a per-app feature. It should be yours, and it should follow you.
The problem with app-locked memory
Here's what happens today. You spend six months teaching ChatGPT your taste, your projects, your recurring questions. It gets good. Then you try Claude for a specific task. Claude knows nothing. Then you try Cursor for code. Cursor knows nothing. The work you did building context in one place is worthless everywhere else.
This isn't an accident. It's the business model. The lock-in is the feature. If your memory lived somewhere portable, you'd leave. So the memory doesn't live somewhere portable.
The cost to you is invisible but real. You re-explain. You re-onboard. You accept a worse experience in a better tool because the better tool doesn't know you yet. You stay in the worse tool because starting over is exhausting.
What a portable memory actually is
Eigen Mesh is a memory that lives in one place you control, and that any tool can read — through a protocol called MCP. Model Context Protocol. It's the part that makes this not-a-notes-app.
Here's the shape. You capture things in Eigen Mesh — your projects, your preferences, your recurring context, the things you keep re-typing. You say it, you meant it, it's saved. Then, when you're in Cursor and you ask "what's the deal with the Berlin client?", Cursor reads your memory and answers. When you're in Claude and you ask "remind me what I decided about the pricing tier", Claude reads the same memory. When you switch to whatever comes next, the memory goes with you.
You explain yourself once. Then you stop.
A concrete week
Monday. You're in Cursor, building a feature. You ask: "What did I decide about the auth flow last week?" Eigen Mesh answers — because Cursor reads it via MCP. You don't open a notes app. You don't scroll a chat history. You ask, in the place you're already working, and your memory answers.
Tuesday. You're in Claude, drafting a tricky email to a client. You ask: "What do I know about how this client likes to be talked to?" Same memory. Same answer. You don't re-explain that they hate exclamation points and want bullet points. The memory already has it.
Wednesday. You're trying a new tool — let's say it's an AI research assistant you just heard about. It supports MCP. You point it at Eigen Mesh. It knows you immediately. No onboarding. No "tell me about yourself." The memory is already there.
Thursday. You decide you don't like the new tool. You stop using it. The memory didn't move. It's still in Eigen Mesh. Nothing was lost.
Friday. You're back in Cursor. You ask: "What's the context on the side project?" Your memory answers. The same memory that answered the new tool on Wednesday. Nothing duplicated. Nothing lost.
This is what portability actually feels like. Not a sync feature. Not an export button. A memory that's just there, in every room you walk into.
Why this isn't just convenience
There's a quieter reason this matters. When memory is portable, you stop being loyal to a tool out of dependency. You can pick the best tool for the job — the best model, the best interface, the best price — without paying the context tax. You can leave a tool that gets worse. You can try a tool that gets better.
Right now, most people can't do this. The memory tax keeps them stuck. Eigen Mesh is the bet that you shouldn't have to pay it.
The ownership part
The memory is yours in the literal sense. Apache 2.0, open source, github.com/hikaru90/eigen. Self-host it with Docker Compose if you'd rather run it on your own hardware. Or use the hosted version — EU servers, no training on your data, €3.99 a month after 100 free credits.
Either way, you can export. You can fork. You can walk away. The memory doesn't belong to us. It belongs to you, and the tools come to it.
That's the whole pitch. Stop re-explaining yourself. Start continuing.
If you're tired of the onboarding loop
If you've ever opened a new AI tool and felt the small dread of "I have to explain everything again" — you don't have to. Not anymore.
Get early access to Eigen Mesh. Capture your context once. See what it feels like when every tool you open already knows you.