Spark
$5/month
For one role, done well.
1 brain.
Core connectors.
Nightly sync.
Nightly sync.
Works with Claude and ChatGPT.
For the people who live in Claude or ChatGPT every day.
hippocampOS fixes that. It remembers your work, your projects, and your context, and hands it to Claude or ChatGPT the moment you ask. No more re-explaining who you are.
You open Claude to draft an email.
But first, you explain your job. Again.
You paste in the background. Again.
You re-upload the file from last Tuesday. Again.
The AI is brilliant. It just has no memory of you. Every chat starts from zero, and you're the one doing the remembering.
If you use Claude or ChatGPT for real work, that's 20 to 30 minutes a day spent rebuilding context. That's the tax. Most people never notice they're paying it.
You feed it your world once: your email, your meeting notes, your files, your documents. It organizes all of it into a private, searchable memory that Claude and ChatGPT can call on whenever you ask.
Then, in any chat, you just say:
"Use my work brain."
Your AI instantly knows your projects, your people, your history. You skip the explaining and go straight to the work.
Got it. Pulling from 14 related threads: the renewal is the Meridian contract at $48k/yr, and Priya flagged pricing concerns on May 28. Here's a draft that answers her two open questions...
Your life isn't one thing. You have a job. Maybe a side project. Maybe research, or school, or a nonprofit.
hippocampOS lets you build a separate brain for each. Your work brain knows work. Your personal brain knows home. They never mix, and you choose which one your AI sees.
One work brain holding your role, your team, your projects, and the last six months of decisions. Fed by your Gmail and your meeting notes.
"Use my work brain. Draft the summary for Thursday's leadership sync."
It already knows: your goals for the quarter, who the stakeholders are, and what shipped last sprint. You edit instead of explaining.
One business brain holding every client thread, your pricing, your product, and the voice you write in.
"Use my business brain. Write a follow-up to the client who went quiet two weeks ago."
It already knows: the whole conversation history, what you quoted, and how you talk. You edit instead of explaining.
A brain per client. The deck history, the stakeholder map, the open risks. Context never bleeds between engagements.
"Use my Acme brain. Prep me for tomorrow's steering committee."
It already knows: every meeting note and every deliverable for that client, and nothing from any other. You edit instead of explaining.
One thesis brain holding the syllabus, the readings, your advisor's notes, and every draft you've written.
"Use my thesis brain. Where does my argument contradict the sources from chapter two?"
It already knows: every paper you've saved and every version you've drafted. You edit instead of explaining.
One home brain for the life admin: travel plans, family logistics, the contractor emails, the receipts.
"Use my home brain. What did the contractor quote us in March?"
It already knows: the quote, the follow-up, and what you agreed to. You edit instead of explaining.
That's it. That's the product.
We tested it against the standard research benchmark for AI memory. The short version:
And it's fast. Under one second from asking to having your full context in the chat. The same lookup through your raw inbox takes the AI two minutes or more.
| Benchmark | R@1 | R@5 | R@10 | MRR | Median | P95 |
|---|---|---|---|---|---|---|
| LoCoMo | 67.07% | 90.08% | 94.71% | 0.7676 | <1 ms | <1 ms |
| LongMemEval-S (scoped) | Not reported | 97.00% | 98.40% | 0.9340 | Not reported | Not reported |
| ConvoMem sample | 39.89% | 58.36% | 65.62% | 0.4790 | 96 ms | 129 ms |
R@K reports whether labelled evidence appears in the first K results. MRR rewards placing the first correct result nearer the top. These are the latest measured values from the public run on July 23, 2026. Read the full methodology and comparisons.
You are handing this product your working life. So here are the rules we operate by, in plain English:
We never train AI models on it. Not ours, not anyone's.
In storage and in transit. There's no version of this product where it isn't.
When you ask a question, we send the relevant slice of memory. Never your whole archive.
Every source is connected by you, filtered by you, removable by you.
Remove a source, a brain, or your entire account whenever you want. Export everything first if you like.
No ads. No data brokers. That's the whole business model.
For one role, done well.
Keep work and life separate
For the true multi-hat operator
Early member pricing, held for the first waves. Works with free-tier Claude and ChatGPT. You pay your AI provider for intelligence. You pay us for memory.
No. Even the free tiers work. hippocampOS does the memory work on our side.
No. Never. It exists for one purpose: answering you.
Export everything as plain files, then delete it all. We keep nothing.
We onboard in waves. Join the list, or skip it with an invite code if a friend has one.
Join the waitlist. We'll write when your wave opens.