Skip to content

AI memory, answered.

Everything people ask before trusting aiperson with the AI they've shaped — ownership, portability, privacy, and getting started.

The basics

What is aiperson?

aiperson is an AI memory layer: a portable, local-first store of who you are and how your AI behaves. The persona you've shaped — voice, values, conventions, refusals, relational history — lives in a signed file you own and is projected into every AI you use, from ChatGPT and Claude to local models. Cultivate one AI, keep it yours.

What is an AI memory layer?

An AI memory layer is a single, portable store of your context and your AI's character that any assistant can read, instead of each app keeping a separate, vendor-locked memory. It holds your memory once, in a place you control, and lets every AI read from it.

What does it cost?

It is free while we build toward our first 100 Founding Members. No credit card required. Sign in with GitHub, run two commands, and your existing AI tools gain a shared memory in under ten minutes.

Which AI tools does it work with?

Thirteen editors and CLIs over MCP — Claude Code, Cursor, VS Code, GitHub Copilot, Windsurf, Zed, JetBrains AI, Antigravity, Augment, Continue.dev, Cline, Aider and Gemini CLI — plus 21 cloud chat surfaces via the browser extension, including ChatGPT, Claude, Gemini, Microsoft Copilot, Grok, Meta AI, Perplexity, Mistral, and the Chinese frontier (Kimi, DeepSeek, Qwen and more).

Ownership and portability

How is this different from ChatGPT memory or Claude memory?

Vendor memory lives on the vendor's servers, in their format, and works only in their app. The March-2026 import features let you copy it out once, lossily. aiperson keeps one persona you hold, signed by your own key, present across every AI at the same time — not a one-time copy. You can export it any time in an open format.

Can I really move my AI to a different model?

Yes — that's the whole point. Cultivate an AI under Claude over a week, then point a local Llama (or ChatGPT, or Gemini) at the same persona file. It greets you the same way, holds the same refusals, and remembers what you were working on. The persona moves intact because it's a file you own, not a setting inside someone else's product.

Can I export everything and leave?

Any time. personkit export --format=markdown emits your whole corpus as Obsidian-ready markdown files, one per thread. Signed JSONL snapshots round-trip byte-for-byte. The format is open, so you can walk away from us entirely and keep everything.

What's the difference between the persona and the corpus?

The persona is who your AI is with you — voice, values, conventions, refusals, signatures — in a signed git-backed document. The corpus is your operational memory — git commits, plans, sessions, captured conversations — in a local SQLite database with hybrid search. Together they make your AI both recognise you and remember your work.

Privacy and trust

Where does my data live?

On your machine. The persona is a git repo you own; the corpus is a local SQLite file at ~/.dotperson/corpus.db. Your local corpus stays there. The relay handles sign-in, GitHub brokerage and synthesis — and it does hold what the capture surfaces you switch on send it, which is how your memory reaches your other devices.

Is aiperson watching everything I type?

No. Capture is opt-in per surface and off by default. There's a master kill switch, a per-conversation skip toggle in the page, a regex redactor that runs before anything leaves the page, and per-surface health dots so you always know what's being recorded.

What can't aiperson see?

Your local corpus database, unless you turn on mirroring for a specific data kind; your Ed25519 secret key, which is generated locally and never transmitted; anything captured while you're signed out; and anything outside the data kinds you've explicitly wired. Email, SMS, notes, calendar descriptions and screenshots are never mirrored by default.

How are my files protected?

Your persona is signed by an Ed25519 key generated on your machine. Every .person.json commit and every corpus snapshot is signed, so tampering is detectable and round-trip recovery is provable. Harvested payloads pass through a secret redactor that strips API keys, tokens and PEM blocks before any write.

Getting started

How do I set it up?

Sign in with GitHub, then run two commands. personkit import synthesises a persona from the context files you've already tuned (CLAUDE.md, Cursor rules, Copilot instructions); personkit setup detects every supported AI tool on your machine and wires them all. Under ten minutes, start to finish.

Do I have to give up the AI I already use?

No. aiperson sits above your assistants, not in place of them. Keep using Claude, ChatGPT, Cursor or a local model — aiperson is the shared memory beneath all of them.

Can I run it myself?

Partly, today. The daemon, corpus and MCP server run entirely on your machine with no relay at all — that is a real local-only mode you can use now. The CLI, corpus and MCP server are source-available under BSL 1.1. The relay's source is not public yet, so you cannot stand up your own relay; a self-hosting guide for it is planned.

About the project

What is the .person Protocol?

The .person Protocol is an open standard for memory, identity, and rights. It is held as a public good by the .person Foundation, a US 501(c)(3) non-profit with an irrevocable charter that forbids conversion to a for-profit. dotPerson PBC, the company you are buying from, is a separate commercial entity that builds and operates products on the open Protocol under a licence from the Foundation. Learn more at dotperson.org.

Who is behind aiperson?

aiperson is built by dotPerson PBC, a US Public Benefit Corporation, as the first commercial implementation of the .person Protocol. Founder Humphrey Theodore K. Ng'ambi (TK) is a systems architect and cybersecurity engineer based across Johannesburg and Lusaka. The project is open about its structure — the full corpus is published.

Ready to keep your AI yours?