Your Personal AI Mind Archive.
Own your AI memory. Simple, private, and yours.
Mind Archive is a local-first application for preserving, organising, searching and exporting your AI conversations, memories, knowledge and files — so that they stay yours, and stay useful even when you change AI provider.
Your archive lives on your own computer. There is no account, no server, and nothing is uploaded anywhere unless you explicitly configure it.
Status: early development. You can import a ChatGPT export, then browse, search and read it. Organising it with tags and projects is next.
The supported way to run Mind Archive is Docker.
pythonorpython3? On Windows it ispython. On Linux, WSL and macOS it is usuallypython3— Ubuntu ships no barepython. Everypython ...command below works either way; use whichever your shell has.
git clone https://github.com/narainsagar/mindarchive.git
cd mindarchive
cp .env.example .env
python scripts/dev.py up --buildThen open:
- Mind Archive — http://localhost:5173
- API — http://localhost:8000
- API documentation — http://localhost:8000/docs
To stop: python scripts/dev.py down. Your archive stays in ./data.
Plain docker compose up --build works too — scripts/dev.py is a convenience
wrapper, not a requirement.
You need Python 3.11 or newer and Node.js 20 or newer.
Backend:
cd apps/api
python -m venv .venv
source .venv/bin/activate # Windows: .venv\Scripts\activate
pip install -e ".[dev]"
uvicorn mind_archive.main:app --reload --port 8000Frontend, in a second terminal:
cd apps/web
npm install
npm run devFull instructions, including WSL2 and Windows specifics, are in docs/DEVELOPMENT.md.
In ChatGPT: Settings → Data controls → Export data. Two things catch people out, and neither is obvious:
- ChatGPT says preparing it can take a few days. Request it before you need it.
- The download link expires 24 hours after the email arrives, and asking again cancels your previous request. Ask once, then wait.
When it arrives, save the .zip into data/inbox/ and Mind Archive imports it
on its own. Or choose the file in the interface — whichever you prefer.
Already keep your exports somewhere? Point the inbox at that folder with
MIND_ARCHIVE_INBOX_DIR; your files stay exactly where they are.
Re-importing later costs nothing: Mind Archive reports what is genuinely new and leaves everything else untouched.
In a hurry? docs/FASTER_IMPORT.md describes an optional script you run in your own browser to get your conversations in minutes. Mind Archive never sees your session and never contacts OpenAI.
Local first. Everything works with no network connection and no account.
Privacy first. Nothing leaves your machine. Cloud features are opt-in, off by default, and can never be enabled silently.
You own your data. Conversations are stored as ordinary Markdown and JSON files you can read, search, back up and move with any tool you like. SQLite is used only for indexing — delete it and your archive is still intact.
No provider lock-in. ChatGPT is simply the first import format. Claude, Gemini, Google AI Studio, local models and future providers all plug in through the same adapter interface. None of them sits at the centre.
Human readable. The interface, the documentation and the stored data are all written to be read by a person.
mindarchive/
├── apps/
│ ├── api/ Python + FastAPI backend
│ └── web/ React + TypeScript + Vite frontend
├── docs/ Documentation
│ └── project-memory/ Decisions, state, milestones, sessions
├── scripts/ Development helpers, and the optional browser script
├── prompts/ The founding specification and AI coding guidance
├── data/
│ ├── archive/ Your conversations. Never committed.
│ └── inbox/ Drop exports here and they import themselves.
├── AGENTS.md Instructions for AI agents and contributors
└── docker-compose.yml
Copy .env.example to .env and edit it. Every setting has a safe default;
the file documents what each one does. .env is git-ignored and must never be
committed.
| Document | What it covers |
|---|---|
| PRODUCT.md | What Mind Archive is and who it is for |
| FASTER_IMPORT.md | Getting your conversations in sooner |
| ARCHITECTURE.md | How the system is put together |
| DEVELOPMENT.md | Setting up and working on the project |
| SECURITY.md | Privacy model and reporting a vulnerability |
| ROADMAP.md | Where the project is going |
| BACKLOG.md | Ideas and known gaps |
| DECISIONS.md | Why the project is built this way |
| GITHUB_PAGES.md | Publishing the documentation site |
Please read CONTRIBUTING.md and AGENTS.md — the second one applies whether you are a human or an AI coding agent.
Outside contributions are not being accepted at this stage. Issues and ideas are welcome in the meantime.
You should be able to clone, install, run, understand and test the project without any private credentials.
Mind Archive is developed with AI-assisted engineering workflows alongside human design, review, testing and decision-making. Every change is reviewed and verified by a person before it lands.
Mind Archive is proprietary software. All rights reserved.
It is not open source and it is not source-available. No licence to use, copy, modify or distribute it is granted — see LICENSE, and LICENSING.md for the plain-language version.
Your archive is yours regardless. Licensing governs the software, never your data.