Building Tabula: One Memory, Every AI

Every time you switch AI tools, you start over. You introduce yourself again, explain what you are working on again, and re-teach the same preferences you taught the last assistant. Whatever context you built up stays locked inside the tool you built it in.

Tabula exists to fix that. It gives you one memory that you own, and it works across every AI you use. Save something in one assistant, and recall it in the next.

Your AI memory is trapped

Most AI memory today is built per app and owned by the company that made it. What ChatGPT remembers stays in ChatGPT. What Claude remembers stays in Claude. Switch tools and none of it follows you. Even when an assistant remembers you well, that memory belongs to the product, not to you, and you cannot take it anywhere else.

That is backwards. The context you build by talking to AI, who you are, how you work, what you are building, is some of the most useful information you have. It should belong to you, and it should work everywhere.

Memory you own, not memory you rent

Your AI memory is the set of facts, preferences, and context an assistant keeps about you. Tabula treats that memory as yours. You decide what goes into it. You can read, edit, export, or delete any of it. And it follows you from one AI to the next instead of resetting every time you switch.

This only works if it stays deliberate. Tabula does not quietly record everything you say. It keeps what you or your assistant decide is worth keeping, because a second brain full of noise is just noise.

Why this is possible now: MCP

A portable memory can exist today because of a standard called MCP.

The Model Context Protocol (MCP) is an open standard that lets AI assistants connect to outside tools and data through one shared interface. It was introduced in late 2024, adopted across the major assistants (Claude, ChatGPT, Gemini, Copilot) through 2025, and donated to the Linux Foundation in December 2025. No single company owns it.

That last part is the whole point. Because MCP is open and the major AIs already speak it, Tabula does not have to talk the industry into supporting it. The connection already exists. Every assistant that supports MCP can read and write your Tabula memory, and every new tool that adopts the standard is one more place your memory works, with nothing for you to do.

How Tabula works

You connect Tabula once, then use it through whichever AI you already open. The setup guide walks through it for each tool.

• Start a plan in one assistant and finish it in another, without re-explaining the background.
• Tell one AI how you like to write, and every AI you connect writes that way.
• Keep your project context in one place, so any assistant you open already knows what you are working on.

Save in one tool. Recall in the next. That is the whole loop.

Your memory is not a product

A memory is only worth keeping if you trust where it lives, so the rules are simple. We do not train on your memories. We do not sell them. We never hand them to outside AI companies. You can see everything stored, change it, take it with you, or delete it for good.

Where this goes

More assistants adopt MCP every month. As they do, your Tabula memory becomes more useful on its own, with no new work from you. That is the bet: instead of every AI building its own walled-off memory, you keep one memory that you own and carry into all of them.

It is early, and we are building in the open. But the goal is worth saying plainly: one memory that is yours, in every AI you use.