“Why does it think that?”
A memory without a source is hard to trust.
See the evidencePersonal context. On your terms.
Your AI’s memory.
Your say.
Inspect it. Correct it. Take it anywhere.
Turn your chats, notes, and work history into memory you can review. Give your assistant the right context, and keep the rest to yourself.
- ✓ Source-backed
- ✓ Permission-led
- ✓ Yours to export
A little context. A better answer.Product demo · example memories
Work space
You
What should I focus on this week?
PersonaKit4 memories shared
Make room for the work that moves things forward.
Every read is saved to your history: which assistant asked, when, and exactly which memories it received.
No account connected
You decide what gets through.
For the AI tools you already use
- Claude
- ChatGPT
- Gemini
- Codex
- Claude Code
Hermes
Your reviewed context, across assistants and agents. Explore compatibility ↗
The missing piece
Being remembered is useful.
Being understood is better.
Your context changes. Your tools change.
You should have a say in what follows you.
“That was me last year.”
An old project or passing preference shouldn’t define you forever.
Keep context current“Do I have to start over?”
Switching assistants shouldn’t mean explaining yourself from scratch.
Take your memory with youNothing about you, without you
A memory.
With the receipts.
Open any memory and see where it came from. Correct the interpretation. Decide where it belongs.
MemoryApproved
You prefer direct, practical feedback over encouragement-first answers.
Try a decision. It resets when you reload.
Why PersonaKit thinks this
“Skip the pep talk. Give me the practical next step.”
“If you disagree, tell me why.”
- Use in
- Work planning
- Sensitivity
- Standard
- Shared with
A conversation is evidence. An AI’s assumption isn’t.01 Your words02 Your decision03 Your boundaries
From your history to your next hello
A little setup.
A better starting point.
- 01Step
Bring a source.
Start with a chat export, notes, or work history you choose.
- 02Step
Review memories.
Check what PersonaKit learned, in groups or one by one. Anything you skip can still be used, marked as not yet checked.
- 03Step
Connect an assistant.
Start with an MCP-compatible pilot or bring a reviewed Markdown export.
- 04Step
Ask. See why.
Get relevant context with sources and a record of what was shared.
Built to grow with you
You change. Your memory should, too.
At first it checks with you. As it learns what you keep and what you correct, it asks less, and still asks when something needs your judgment.
- You work
- PersonaKit learns
- You stay in control
The PersonaKit principle
The product isn’t the memory.
It’s your control over it.
Your store, in an account you own.Your sources, attached to your memories.Your choice, every time context travels.
Your context goes with you
New AI tool.
Still you.
Take your memory to any assistant.
Connect it, or export a copy.
Pilot
MCP-compatible assistants
Connect an MCP-compatible assistant to the memories you choose to share.
We’ll guide you through setup
Developer pilot
Claude Code, Codex & CLI tools
Guided MCP setup for your coding workflow, including other CLI tools with MCP support.
We’ll check your tool’s compatibility
Export first
ChatGPT, Gemini & Claude exports
Bring a chat export from ChatGPT, Gemini or Claude today. Direct connectors for ChatGPT and Gemini are planned.
You choose what to upload
Session import
Hermes
Start from exported chats or assistant history you choose to share. Supported imports include Hermes session exports.
New chats? Import a fresh export.
Portable format
Markdown & JSON
Review a memory pack with source quotes before sharing it with another tool.
Markdown (.md) · example memories. Open in any text editor; no signup needed.
Shared files won’t update when you change your memories or permissions.
Try a different question
Don’t take our word.
Ask your memory.
Choose a prompt. Read an example answer right here.
What should I focus on this week?↗
Example answer
Make room for the work that moves things forward.
- 01Protect two mornings for focused work.
- 02Finish the proposal before taking on more.
- 03Keep Friday open for your weekly review.
You prefer a short, practical plan. Start here.
Work only
View in the product demo ↑How should you give me feedback?↗
Example answer
Directly, and with a next step.
- 01Skip the encouragement and say what to do next.
- 02If you disagree with me, tell me why.
That comes from two things you said, in July and August.
Work only
What did you leave out, and why?↗
Example answer
Two memories stayed back for this question.
- 01One is private and outside the work scope.
- 02One is out of date and needs your review.
I know they exist, but I never received what they say.
2 withheld · Work planning
See what was withheld ↑Can you use my work memories in a public bio?↗
Example answer
Public writing needs its own permission check.
- 01No memories are approved for public writing in this demo.
- 02Work-only approval does not grant public-use permission.
- 03Tell me the facts you want to include, or review their scope.
No personal draft generated: public-use context is withheld.
0 shared · 6 withheld · Public writing
What have you read about me?↗
Example history
Every read is saved to your history.
- 01Today, 9:12 · your Work assistant asked a planning question.
- 02It received 4 memories, and 2 were held back.
- 03No other assistant has read anything.
You can see which assistant asked, when, and exactly which memories it received.
A few things, up front
Trust needs
specifics.
What is PersonaKit?+
PersonaKit is a personal AI memory control layer for people who use AI every day. It turns selected chats, notes, and work history into source-backed memories you can review, correct, restrict, and export to the assistants you already use.
The private pilot starts with guided setup. The memories shown on this page are examples.
Where does my memory live?+
PersonaKit is designed to keep your memory store and read log in an account you own. Processing services handle the context needed for a request.
This website cannot access your personal memory. Before you share a source, we’ll explain the storage setup, providers, and what they can access.
Do you train on my conversations?+
No. PersonaKit’s policy is not to train models on your conversations.
Connected assistants and processing providers have their own terms and settings. We review those with you during setup before you share a source.
Can I correct or remove a memory?+
Review decisions let you edit, restrict, or reject records. Restrictions affect future PersonaKit reads. A previous export or an assistant’s existing conversation cannot be recalled by changing a record.
Do I have to check every memory?+
No. Anything you skip can still be used and is marked as not yet checked. Today, only you can confirm a memory. Over time we aim to ask you less, by learning from what you keep and correct. If PersonaKit ever starts confirming routine memories for you, we’ll tell you first and you’ll be able to turn it off.
Does it capture everything I say?+
No. Capture depends on the integration and what you choose to share. A connector is not automatic access to your entire conversation history. The pilot will make the boundary visible for your setup.
What is real in this demo?+
The memories shown on this page are examples, and the answers are scripted. Exploring the demo makes no model calls and does not connect your accounts. The private pilot starts with guided setup. The waitlist form is a local preview; it does not submit details, download a draft, or reserve a place.
An assistant by PersonaKit
Meet Trinity.
Help that gets to work.
Trinity is the assistant that uses your memory. It learns how you work, gets the groundwork done before you ask, and checks with you before anything leaves your hands. Invitation only.
Early access. A personal invitation.
Join the
private waitlist.
A small group at a time, for PersonaKit memory and for Trinity.
- Your email, not your history
- Personal follow-up
- Leave whenever you like
We’re getting ready to welcome you. For now, the form is a preview - not a signup.