Model-neutral · MCP-native · EU-hosted
Memory for the model you already pay for.
Memexa is a personal knowledge graph that every MCP client can read and write. Save a thought in Cursor, recall it in Claude — one brain, owned by you, hosted in the EU.
Free plan · no credit card
One memory, every client
Tell it once. Ask anywhere.
Your AI clients share a single memory over MCP. What you save in Cursor on Monday, Claude can recall on Friday — with the source.
Example: a decision saved from Cursor on Monday is recalled by Claude Desktop on Friday, with the source note cited.
Scripted illustration of real MCP tool calls.
How it works
One URL. Any client.
Connect
Paste your personal MCP endpoint into any client. Sign in once in the browser — no tokens to copy.
claude mcp add --transport http --scope user memexa https://brain.memexa.ai/mcpFree — works on every Claude plan.
Work as usual
Your client saves decisions, notes, and findings as you go. Memexa links them into one graph of topics, people, and projects.
Ask anywhere
Any connected client searches your brain and answers with sources — across sessions and across tools.
Why bring your own client
Don't pay for intelligence twice.
Most "AI memory" products bundle their own model and charge you for it. Memexa doesn't. It gives one shared, inspectable memory to the AI subscriptions you already have.
| Comparison criteria | Memexa | ChatGPT Memory | Notion AI | Obsidian |
|---|---|---|---|---|
| Works across AI clients | Any MCP client | ChatGPT only | Notion only | Via plugins |
| Ships its own model | No — bring your own | Bundled | Bundled | None |
| See & edit every memory | Full notes UI + graph | Limited list in settings | Yes, inside Notion | Yes, local files |
| Open Markdown export | Anytime, one click | Account data export (JSON, not Markdown) | Workspace export | Native — it's files |
| Hosting | EU, GDPR | US (OpenAI) | US by default | Your device |
Swipe →
Based on publicly documented behavior, July 2026.
Ownership & privacy
Your data stays yours.
Hosted in the EU under the GDPR. Export all your notes as open Markdown, anytime. Nothing you store here trains a model.
EU / GDPR hosting
Your content is stored and processed in EU data centers. A few service providers transfer metadata to third countries — safeguarded by Standard Contractual Clauses (SCCs). Your content is not affected.
Export anytime
One click downloads all your notes as open Markdown files. No lock-in, no dead end.
No training
Your notes and documents never train a model — not ours, not anyone's.
Pricing
Start free. Upgrade when it grows.
Yearly billing gets two months free — and Pro costs less than one extra AI subscription, because Memexa doesn't ship one.
Free
For getting started and small graphs.
- ~150 notes
- 50 document pages
- 1 AI client
- 20 questions / day
- 1 email capture sender
- Semantic search + graph
- Export anytime
- EU-hosted
Pro
Your daily memory, without limits.
- Unlimited notes
- 500 document pages / month
- Unlimited AI clients
- 200 questions / day
- 2 email capture senders
- Everything in Free
Power
For heavy users.
- 3,000 document pages / month
- 1,000 questions / day
- 3 email capture senders
- Priority processing
- Everything in Pro
A note from the founder
I built Memexa because I kept re-explaining the same context to every AI tool I use — my projects, my decisions, the people I work with — and losing it with every new session.
The honest state of things: Memexa is young, and you won't find a wall of testimonials here — I'd rather earn them. What you get today is a working shared memory over MCP, an app where you can see and edit everything your clients store, and an export button that always works.
If that sounds useful, try the free plan — and tell me what's missing. I read every reply.
— Thomas, founder · hello@memexa.ai
FAQ
Common questions
What is MCP — and do I need to understand it?+
MCP (Model Context Protocol) is an open standard that lets AI clients connect to external tools and data. If your client supports it — Claude, ChatGPT, Cursor, and many local setups do — you add your Memexa URL once and you're done. You never touch the protocol itself.
How is this different from ChatGPT's built-in memory?+
ChatGPT's memory lives inside ChatGPT: one vendor, a limited list in settings, no export. Memexa is one memory that every MCP client shares — what Cursor saves, Claude can cite. Every note is visible in the app, editable, linked into a graph, and exportable as Markdown at any time.
Does it work with Cursor?+
Yes. Cursor speaks MCP — add the Memexa endpoint to ~/.cursor/mcp.json and sign in once in the browser. Claude Code, Claude Desktop, ChatGPT, and other MCP clients connect the same way.
Which AI model am I locked into?+
None. Memexa ships no model of its own and is deliberately model-neutral. You use the client and model you already pay for — and can switch anytime without losing your graph.
Where is my data stored?+
Your content — notes, documents, embeddings — is stored and processed in EU data centers under the GDPR. A few service providers (e.g. email delivery, payments, error diagnostics) may process metadata in third countries, safeguarded by Standard Contractual Clauses; your content is not affected. Details are in the privacy policy (German).
Can I get my data out?+
Anytime. One click exports all your notes as open Markdown files in a ZIP. There is no lock-in — the data is and stays yours.
Is my content used to train models?+
No. Your notes and documents are never used to train models — not by us and not by third parties.
The idea
An idea from 1945, finally practical.
In his 1945 essay As We May Think, Vannevar Bush imagined the Memex: a desk that stores everything you know and links it by association — the way thinking actually works. The essay went on to inspire hypertext and the web, but the personal memory device was never built. It needed a machine that could read, link, and recall on its own. That machine is the AI model you already use.
“An enlarged intimate supplement to his memory.”