Secure.
Simple.
Seamless.
Smart.
One private memory and notes vault for every AI you use — secure enough to trust with what matters, simple enough to forget it's there, seamless across every tool you already use, and smart enough to surface the right thing at the right moment. Your AI does the thinking. Penny does the remembering.
First 250 accounts join as founding members - $59 a year, locked for life.
Trust you can check.
Most products ask you to trust a privacy policy. Every time your memory is unlocked, it's recorded — and you can see the full list in your private dashboard. Once a day that record is sealed into a public, tamper-evident log, so trust isn't something we ask for. It's something you can verify.
Instead of having your data go into a black box, you can log into your Penny dashboard and see every piece of info it's tracking, AND every time it's been accessed.
Encrypted at rest, with a key unique to you
Every memory is sealed with AES-256-GCM under a 256-bit key unique to you — and the keys live in a separate, hardware-backed cloud account from the database that holds your data. A breach yields ciphertext and nothing else.
No cross-account leaks
Each record is cryptographically bound to your account and its exact location. It can't be moved into anyone else's memory — the math simply fails.
A public, tamper-evident access log
Every decryption of your memory is recorded before the key is used. See every access in your dashboard; each day's record is sealed into an append-only public log hosted on a different cloud than your data, and your dashboard checks it against the log for you. Out-of-the-ordinary access by a person is published individually, the moment it happens.
Nothing is ever silently gone
Delete a memory and it moves to a 30-day recoverable trash. No agent — even a malfunctioning one — can erase what matters before you can undo it.
Secrets never make it in
If a password, API key, or token slips into a conversation, Penny strips it out before anything is stored. An accidental paste never becomes a permanent memory.
Nothing is captured behind your back
Penny saves memories from the conversations and tools you've connected - never by watching your screen or recording in the background. What goes in is what you chose to connect, and every read of it is on a record the public log seals.
Set it up in one conversation. Then forget it's there.
No dashboards to configure, no prompt engineering, no manual. The first time you use Penny, your AI offers to set it up — a handful of questions, one short snippet to paste into your AI app, done.
From then on it keeps notes for you, silently. You never say "remember this" — it captures the decisions and preferences worth keeping as you work, and recall happens on its own. Set a preference once and every connected tool inherits it.
It took 30 seconds to set it up, then I almost forgot it was there — until my AI reminded me about something out of the blue. It was like magic.
Up and running in four steps
Connect once
Authorize Penny in your AI tool with a single sign-in over MCP. No keys to paste, no install to babysit.
Just talk
Use your AI normally. Key moments become durable memories automatically — no "save this" required.
Recall everywhere
Open any other connected tool and it already knows you. Your memory follows you, not your model.
It gets sharper
A background pass consolidates what matters and refines your profile, so recall keeps improving the longer it knows you.
One memory. Every tool. Already there when you open the next one.
Penny doesn't live inside one chatbot. It lives at the connection layer between you and all of them — so the same memory rides along everywhere, and you stop copy-pasting context between tools for good.
Your coding subconscious. The Claude Code, Cursor and Codex plugins inject your preferences and the project's accumulated memory at the start of every session — the project recognized automatically from your repo, zero configuration. Yesterday's decisions are simply present before you type a word.
Open Cursor tomorrow morning and it already knows what you were building yesterday in Claude Code, or switch tools mid-project and the full context is just there. Like magic.
One connection, every AI
Penny speaks an open standard — the Model Context Protocol — so it works across Claude, ChatGPT, Perplexity, Cursor, Codex, and Claude Code, and any tool that adopts MCP next. No per-app plug-ins.
Richer while you're away
When a session ends, its secret-scrubbed transcript feeds background processing that distills what was learned into that project’s living memory for next time.
Disconnect when you want
Cut off any tools access from one dashboard. Your memory lives in your account, not inside your AI tools. Read, edit, delete when you want, and export is just a click away.
Import isn't a one-way street
Plenty of products can pull your memory in (easy in but hard to get out). Yours leaves as easily as it arrives: read, edit, and delete from your dashboard, and even export. Leaving is a feature.
Memory that thinks before it answers.
An AI is only as good as what it can find. Penny's whole job is to surface the right thing at the right moment — even when you ask imperfectly. It searches by meaning and by exact words at once, follows a knowledge graph built from how you actually work, and consolidates everything in the background so your memory gets cleaner over time, not noisier.
Your AI now finds the right memory even if you phrase a question completely differently. Penny reviewed last night’s conversation while you slept, keeping only what was actually new.
A short, always-on set of core memories rides in every conversation; everything else lives in a deep notes archive, pulled in only when relevant. The essentials make it feel like it already knows you. The archive means nothing useful is ever out of reach.
Every query runs semantic and keyword search at once, then fuses the results — so you never lose a note because you phrased the question differently than you wrote it.
After you work, background processing decides what's genuinely new versus already known, writes only what matters, then merges duplicates and resolves contradictions so memory stays clean.
Results are scored by a single transparent, ablation-tested formula — relevance, graph proximity, recency, and the importance you've signaled — so the right note wins even when it isn't the newest.
Structured trackers (habits, metrics, goals) and recurring rhythms (check-ins, reviews) add accountability over time — not just notes and memories. Penny can even manage your to-do list.
Why a dedicated memory layer?
| Feature | Built-in chatbot memory | Penny |
|---|---|---|
| Works across different AI tools | No | Yes |
| Take it with you when you switch tools | No | Yes |
| Encrypted at rest with a per-user key | No | Yes |
| Tamper-evident public access log | No | Yes |
| You choose which tools can read it | No | Yes |
| See and edit exactly what's stored | Limited | Yes |
| Recall by meaning, keyword, and relationships | Basic | Yes |
| Same memory in your coding tools | No | Yes |
| Improves itself in the background | No | Yes |
| Keeps a history — never silently overwritten | No | Yes |
| 30-day undo on every deletion | No | Yes |
| Trackers for habits, metrics, and goals | No | Yes |
| Strips out secrets before storing | No | Yes |
| Doesn't keep your raw chat logs | No | Yes |
| Memory leaves as easily as it arrives | No | Yes |
Questions and answers
Is my memory end-to-end encrypted?
No — and we're precise about that. Your AI tools need to read your memory in plain text to use it, so end-to-end encryption isn't compatible with how a memory layer works. What we lock down instead is everything around that moment: your content is encrypted at rest with a 256-bit key unique to you, encrypted in transit, and every decryption is recorded in your dashboard and sealed daily into a public, tamper-evident log. Any access by a person outside routine operation is published individually. We don't claim we can't read your memory; we commit that we won't — and the log means you would know if anyone ever did.
Which AI tools does it work with?
Any tool that speaks the Model Context Protocol. Today that includes Claude on desktop and web, ChatGPT, Perplexity, Cursor, and Codex — with more arriving as MCP adoption grows.
Can I export or delete my data?
You can read, edit, and delete your memories and notes from your dashboard today, and request a full export anytime - we fulfill export requests on request. One-click export in a portable format is on the way. Your memory is yours; we just hold it.
Do you use my memory to train models?
No. Your memory exists to serve you across your own tools. The intelligence is whatever LLM you connect. Penny is the memory layer, not a model. For any background processes we use only directly hosted hardware and zero-data-retention (ZDR) providers so none of your data is available to companies that would use it for training.
Does Penny store my passwords or API keys?
No. Before anything is saved, Penny scans for secrets — passwords, API keys, access tokens — and strips them out. If one slips into a conversation, it never becomes a memory and never shows up in search.
Do you keep my chat logs?
No. Once Penny has distilled the facts worth remembering, the raw text is discarded. What's kept is a short, deliberate record you can read, edit, and delete — not a transcript.
Is Penny open source?
No — the memory engine is proprietary. What we make verifiable instead is your access to your own data: a complete record of every time your memory is unlocked, sealed daily into a public, tamper-evident log, plus full read, edit, and delete from your dashboard. You don't have to take our word for it.
One price. Everything included.
Unlimited memories and notes. Every MCP-capable AI tool. Per-user encryption with a public access log. Search by meaning and by keyword. Edit and delete anytime, export on request. Email support.
| Feature | Most popular Founding member $59/ year | Standard $99/ year |
|---|---|---|
| Everything included | Included | Included |
| MCP-capable AI tools | All supported tools | All supported tools |
| Public access log | Included | Included |
| Export | On request today | On request today |
| Free trial | 14 days, no card | 14 days, no card |
| Rate lock | Locked for life | Standard annual rate |
- Everything included
- Included
- MCP-capable AI tools
- All supported tools
- Public access log
- Included
- Export
- On request today
- Free trial
- 14 days, no card
- Rate lock
- Locked for life
- Everything included
- Included
- MCP-capable AI tools
- All supported tools
- Public access log
- Included
- Export
- On request today
- Free trial
- 14 days, no card
- Rate lock
- Standard annual rate
I'll remember this.
Connect me once, and every tool you use knows what matters — and only you decide what I keep. First 250 accounts: $59 a year, locked for life.