Inputs
paste · global paste · drop · URL · pdf · notes
№ 01Capture
Saving a thought should be as easy as having one.
Paste a thought, a link, or a whole article — or drop in files and URLs. memori turns each capture into a clean, tagged note in your knowledge tree. Try it below: type anything and capture it.
Paste — a thought, a quote, or a whole article, straight into your library
Global paste — capture from any app, no window-switching
URL import — paste a link; the page comes in as a note
PDF → markdown — drag in .pdf, .txt or .md files; PDFs arrive as clean markdown
Auto-tagged — every memory lands in your knowledge tree, connections suggested for review
Live demo — capture feed3 memories
Spacing effect, quantified
Across 254 studies, spaced practice beat massed practice in 259 of 271 comparisons. Retention gains of 10–30% — the single cheapest upgrade to any study routine.
“Memory is the residue of thought.”
“You remember what you think about, not what you want to remember. Design for the thought, and retention follows.”
Active recall beats re-reading
Testing yourself is not measurement — it is the learning event itself. Every retrieval strengthens the path more than another passive pass ever could.
№ 02Organize & search
A memory map that draws itself.
No folders to maintain. memori tags every memory and suggests connections you review and apply — the memory map shows how it all hangs together. Type a word; your past self answers.
3 of 3 memories
Spacing effect, quantified
Across 254 studies, spaced practice beat massed practice in 259 of 271 comparisons. Retention gains of 10–30% — the single cheapest upgrade to any study routine.
2026-08-28
“Memory is the residue of thought.”
You remember what you think about, not what you want to remember. Design for the thought, and retention follows.
2026-08-21
Active recall beats re-reading
Testing yourself is not measurement — it is the learning event itself. Every retrieval strengthens the path more than another passive pass ever could.
2026-08-14
Connection map — shared tags
article
highlight
idea
note
№ 03AI & agents
Chat with your own memory.
Chat answers from your own archive with numbered citations — you see exactly which memories every answer used. Your agents join too: one MCP endpoint for Claude, ChatGPT, Cursor or your own scripts.
Numbered citations — every answer shows exactly which memories it used
Web search included — when your archive doesn't have the answer, on every plan
MCP + hashed API keys — one endpoint your agents can search, add and list
Live demo — retrieval console3 memories online
idle — waiting for a question
Retrieved memories will appear here — sources first, always.
Ask something. I'll show you which memories I used — and cite them.
№ 04Portability
One memory, every agent — yours to keep.
Memory saved inside one assistant stays behind when you switch tools. memori gives every agent one shared memory over a single MCP endpoint — with two-way GitHub sync in plain markdown, and one-click export at any moment.
One memory — every agent
fig. 04a — every agent reads & writes the same library
Any agent, over MCP — Claude, ChatGPT, Cursor, your own scripts
One shared memory — what one agent saves, every agent knows
Switch tools, keep everything — no per-app brain left behind
Backup & sync — automatic
fig. 04b — backups run themselves · sync flows both ways
Export at any moment — one click, your whole account as JSON
Automatic backups — continuous, nothing to remember
2-way sync — your GitHub repo in plain markdown, changes flow both ways
Live export — this page's library3 memories · yours
The file you download is the library you've been using on this page. Capture something in § 01, come back, and export it again.
$ memori export — press “Run export” to generate a real file
an export file will assemble itself here
— no file generated yet
№ 05Principles
Four rules, no exceptions.
01
No lock-in
Your archive is yours to leave. One-click export, GitHub sync in plain markdown, full account deletion — leaving is an export, not an escape.
02
Open formats
Markdown in, Markdown out. Metadata in JSON that any tool from this decade can read.
03
Inspectable AI
If a model used your memory, you get to see exactly which parts — citation by citation.
04
Never training data
Your memory is not our corpus. Full stop.
№ 06Questions
Fair questions, straight answers.
Where does my memory live?
Your memori account runs on managed cloud infrastructure — nothing to host, nothing to lose. And it's portable by design: automatic backups two-way sync to your GitHub repo in plain markdown (Open Knowledge Format), and a one-click export downloads your whole account as JSON.
Which AI models can use it?
memori's chat answers from your archive with numbered citations — you see exactly which memories each answer used. Your agents get the same library over MCP: one endpoint lets Claude, ChatGPT, Cursor or your own scripts search, add and list the same memories.
What can I capture?
Anything text-based: paste a thought, a quote or a whole article; drag in .txt, .md or .pdf files — PDFs arrive as clean markdown; or import a URL. Every memory is auto-tagged into your knowledge tree, and suggested tag connections wait for your review.
What does early access cost?
There's a free plan — 100 memories and 50 chats a month — and Pro at $8 a month with 1,000 memories and 500 chats. Stripe billing, cancel anytime. And the exit stays open: export is one click, and your data is never training data.
Start a memory worth keeping.
Free plan · Pro $8/mo · cancel anytime