A memory for your AI
OwnLore gives ChatGPT, Claude and every other AI a memory that belongs to you. Tell it once — from then on every AI knows it, in every new chat and in every AI tool.
You
Write to my landlord about the heating, please.
Your AI
On it. I'll take the address from your notes and the fault you wrote down on 3 February — including the deadline you set yourself.
read: Flat · Open items
The difference
What should I tackle first this week?
Without OwnLore
I can't tell you — I don't know what you're working on. Tell me more about your projects and I'll happily help you sort them.
With OwnLore
The tax paperwork. You noted the 15th as your deadline, and that's Friday. Then the quote for Meyer — it's been sitting with you since Tuesday. The renovation can wait; you're the one waiting on a reply there.
The difference isn't the AI. The difference is that it knows you.
The obvious question
A little, yes. Except that little belongs to the provider: you barely see what's in it, you can't get it out, and switching to another AI leaves it behind.
OwnLore turns that around. Your memory sits with you — every AI may look inside, none of them owns it.
Built-in memory
OwnLore
Belongs to
the AI provider
you
Inspect it
in fragments
every line, any time
Switch AI
starts from nothing
comes along
Size
a few sticky notes
as much as you like
Where it applies
the chat app only
chat, editor, terminal, agents
Everywhere, not just chat
That bit of built-in memory only exists inside the provider's chat app. The moment an AI works for you somewhere else — Claude Code in your terminal, Codex on your project, Cowork on your tasks, Cursor in your editor — it starts from nothing again. And these tools share nothing with each other anyway.
OwnLore plugs into all of them at once. What you tell the chat in the morning, the agent knows in the afternoon — and whatever it learns is back in your chat that evening.
In chat
ChatGPT
Claude
While coding
Claude Code
Codex
Cursor
Copilot
Windsurf
Zed
While working
Cowork
Claude Desktop
And in whatever comes next: OwnLore rides on the open MCP standard, not on a list of names.
What changes for you
Your projects, your preferences, the names of your people. Write it down once — from then on every AI reads it before it answers.
Claude today, ChatGPT tomorrow, something that doesn't exist yet next year. Your memory moves with you, because it never belonged to the AI.
They are ordinary text files in your own storage. Read them, change them, take them with you — with or without us.
Set up in ten minutes
1
Email and password. The first 30 days cost nothing.*
2
Google Drive, a private GitHub repo or your own server. OwnLore only ever gets the one folder or repository you grant — it never reaches the rest. One click, we do the rest.
3
Paste one address into your AI, confirm once. From then on it thinks along.
Nothing to install, nothing to program.
Your notes stay yours
Your notes do not live with us but in your own storage. We only remember that they exist — not what is in them. Even the file names are encrypted on our side.
Hosted in the EU, under the GDPR.
No profile built from your content. Ad measurement only if you agree.
Your files stay exactly where they are. Afterwards too.
OwnLore saves the previous version before every change. One click brings it back.
Price
€2.99
per month
or €19.99 a year — a good five months on us.
30 days free*
Billing starts only after that. Cancel any time.
* The free trial is available once per brain. If you connect storage that has already had a trial, your subscription starts right away — without the free period, but with everything else.
Final price, no VAT shown (small business rule, section 19 German VAT Act). Card payment via Stripe.
For those who want the detail
This is where it gets technical. You need none of it to use OwnLore.
Works with everything that speaks MCP
Claude
ChatGPT
Cursor
Claude Code
Windsurf
Zed
Codex
Copilot
Cowork
OwnLore is a hosted MCP server (Model Context Protocol) — the interface AI tools use to reach external tools. Anything that speaks MCP can read and write your brain: Claude, ChatGPT, Cursor, Claude Code, Codex and whatever comes next. It connects over a URL and OAuth — no download, no local process.
The usual way to give an AI a knowledge base cuts documents into snippets and turns them into number vectors; on every question the most similar snippets get gathered up anew. Connections are rebuilt from scratch each time and are gone once the answer is written, and contradictions between two documents never surface. OwnLore takes the route that became known as the LLM wiki: the thinking happens once, when something is filed. New findings are worked into existing notes rather than parked beside them, contradictions get resolved or flagged, cross-references added. From the summary in each note's frontmatter OwnLore maintains a table of contents — the AI reads it and opens two or three notes on purpose, instead of searching everything. No embeddings, no vector database, nothing unreadable: what you end up with is a reference work you can page through yourself.
The backend is a folder in your Google Drive, a GitHub repository (private included) or your own server via FTPS or SFTP. The notes are ordinary Markdown files in a folder structure you choose — readable with any editor, with Obsidian, with git.
The default is append, not overwrite. Every change is preceded by a snapshot, deleting is a soft delete into the trash, and editing runs as find-replace with unique-match-or-reject. Guardrails block oversized or destructive writes before they happen.
The dashboard shows which AI tool did what and when — with line and size changes, the outcome and, if a write was rejected, the reason. One click rolls a change back.
Migrating to different storage is built in and stays available after the subscription ends. The entire brain folder is copied including its version history; index and log then move with it. The old storage is merely disconnected — the files stay with you as a backup.
Our database holds no content. File names and paths are AES-256-GCM encrypted; per note we keep only an opaque ID, a hash and the size. We do not store diffs either — they are reconstructed from the snapshots in your storage when you look at them.
In your storage
Every note as Markdown
Every previous version
Names and paths in plain text
In our database
Encrypted references
Hashes instead of content
Opaque IDs in the log
Frequently asked
On its own only to a very limited extent. With OwnLore it can: your notes live in your own storage, and the AI reads them before it answers. Whatever is written down is there again in every new chat — even weeks later.
Built-in memory belongs to the provider. You barely see what is in it, you can't get it out, and switching to another AI leaves it behind. With OwnLore they are ordinary text files in your storage: readable, editable, portable.
Yes. They all read the same memory. What you tell Claude, ChatGPT knows afterwards. Any tool that speaks the open MCP standard is supported — including Claude, ChatGPT, Cursor, Claude Code, Codex, Windsurf and Zed.
In storage you provide yourself: your Google Drive, a private GitHub repository or your own server via FTPS or SFTP. Neither the contents nor the file names sit with us in plain text — only encrypted references and checksums.
No. You create an account, pick a storage location with one click and paste one address into your AI. No download, no terminal, no code.
Yes. OwnLore works with ordinary Markdown files in a folder structure you choose — exactly the format Obsidian, Logseq and any text editor use. If you connect storage that already has notes in it, everything stays as it is.
€2.99 a month or €19.99 a year, with 30 days free before that. One price, no tiers, cancel any time. Final price with no VAT shown (small business rule, section 19 German VAT Act).
Nothing. Your files live in your own storage and stay there untouched. We only block access for the AI tools; the dashboard and migrating to different storage stay open.
Yes — and that is where a memory is missing most. The providers' built-in memory only exists in their chat app; agents in your terminal, your editor or a work tool start from nothing every time and share nothing with each other. OwnLore plugs into all of them at once: the same knowledge in chat, in Claude Code, in Codex, in Cowork, in Cursor.
An LLM wiki. A RAG system cuts your documents into snippets and turns them into numbers — afterwards you cannot read back what was stored. OwnLore keeps plain Markdown notes and has your AI maintain them: new findings get worked in rather than parked beside the old ones, contradictions get resolved, and a table of contents tells the AI which note to open. That is the pattern known as the LLM wiki — only without setting up a coding agent on your own machine: versioned before every change, on all your devices and in every AI tool at once.
30 days free.* Then €2.99 a month. Cancel whenever you like.
* The free trial is available once per brain. If you connect storage that has already had a trial, your subscription starts right away — without the free period, but with everything else.