About
Why “memex”
“Consider a future device … in which an individual stores all his books, records, and communications, and which is mechanized so that it may be consulted with exceeding speed and flexibility. It is an enlarged intimate supplement to his memory.” Vannevar Bush, 1945
The memex was a machine Vannevar Bush described in The Atlantic in July 1945, in an essay called “As We May Think”. It was never built. He imagined a desk where everything you had read and written was kept, linked by trails you made yourself, and could be found again at speed.
Eighty years later the storage is trivial and the finding is solved. What Bush could not have anticipated is that we would end up with machines that read and write alongside us, and that the hard question would become which of them gets to hold the record.
What memex is
A single shared knowledge base across all your chatbots and AI agents. Let them know once and they will never forget.
Memex keeps your notes as plain Markdown, makes them searchable by meaning, and hands them to whichever assistant you are working with. It has no AI of its own. The thinking is done by you and by the assistants you already use and already pay for. Memex makes sure you are all working from the same page, and that the page belongs to you.
Everything in it is a file you can open, correct, roll back, export or delete. Your assistants can read from it and write to it, on terms you set for each of them.
Memex is free, and the accounts that pay the providers behind it are ours. So that a runaway script cannot spend our money on your behalf, there are limits on how much any one knowledge base can make memex buy in an hour and in a day. They are set to be out of your way in ordinary use, and they exist for the same reason an assistant’s writes wait for your review: memex keeps a person in the loop. To step outside them, bring your own API keys in Settings — a section on your key is not limited at all, because the bill is yours — or run memex on your own machine against a model that costs nothing.
Who built it
Memex was built by someone who spent more than two decades looking after data, knowledge and digital governance for people and institutions. Not as a hobby project that went looking for a problem, but out of watching, for a long time, how people actually process information, and how knowledge systems get conceived and then built.
Most of what is here comes from that. The insistence on plain files, the review step in front of anything an agent writes, the refusal to host an AI inside a tool whose job is to hold the record: none of those are technical preferences. They are what you conclude after twenty years of seeing whose hands the knowledge ends up in.
It was built for its author first, and then opened to people who wanted the same thing.
Where it is now
Early Access, invite-only, free. It is not a preview. Everything described on the site works today. Invites go out in batches.