I spend my working day inside AI agents. Code in one, research in another, operations in a third. They are astonishing — and every morning, every one of them meets me for the first time.
The memory I build inside one tool is hostage to that tool. Teach an agent your architecture, your conventions, your hard-won lessons, and all of it lives — and dies — inside that vendor's walls. Switch tools, open a second one, or just start a fresh session, and you are a stranger again, re-explaining things you have already explained a dozen times.
That bothered me more than any other problem in this industry, because of what it implies about ownership. Your memory is the most personal artifact there is. It should not belong to whichever company happens to host this afternoon's conversation.
Memory should live below the agent layer, and it should belong to the user. That is the problem I want to spend the next decade solving. Vertiso Memory (vmem) is the start of it.
What it is
Vertiso Memory is a memory service that sits underneath your AI
tools — platform-independent, device-independent, and owned by you.
Your agents connect to it the way they connect to any tool: over MCP,
a REST API, or the vmem command line. Claude, Codex, Cursor,
whatever comes next — they all read and write the same memory,
because the memory is yours, not theirs.
It is not a notes app, and it is not a pile of transcripts with a search box. When you tell vmem something, it reads what you wrote and breaks it into the facts inside: the decision you made and why, the rule you want followed, the task you committed to, the lesson that cost you something, the project it all belongs to. Those typed facts link into a personal knowledge graph — people, projects, topics — with your original words preserved at the bottom of every chain. Summaries are for finding things; your verbatim is the record.
What it does
A session with a connected agent starts with a handshake. The agent says hello and gets a primer: who you are, how you like to work, what is open, what matters right now. No warm-up interrogation. From there:
- Remember. "We decided to ship Fridays only — remember that." One sentence in a tool you already use, and it is kept, typed, and linked.
- Recall. Ask "what did we decide about the rollout?" and get a synthesized answer with receipts — the actual passages, your actual words, cited underneath. If a fact was corrected later, the correction wins and says what it replaced.
- Rules that follow you. Your preferences and constraints ride into every session, in every tool. Say it once. Every agent you work with afterward already knows.
And because the memory is yours, it behaves like property: inspect it on the web, edit it, export all of it, delete it. No hostages.
Who it's for
Today: people who are wired all the way in. Developers running multiple agents, builders shipping with AI daily, operators running AI-forward businesses — anyone who feels the re-explaining tax every single day and knows exactly what it costs. It is a premium tool for people who work this way now.
Soon: everyone, because everyone is about to work this way. The architecture does not care whether the memory is a deploy convention or a preference your assistant should honor. It is universal on purpose.
The bet
Retrieval is table stakes. The bet behind Vertiso Memory is compounding: every session makes the next one more useful. A month in, your agents know your projects, your people, your rules, and your history — and that only accelerates. Your memory becomes the most valuable tool you own, precisely because it is the one thing no platform can give you and no platform can take away.
That is worth a decade.
Vertiso Memory is live today at memory.vertiso.ai. Connect an agent, tell it something worth remembering, and ask for it back tomorrow — from a different tool, if you like. That moment is the whole product.