Private, offline mind mapping

New in v1.19.0: Use Claude from Anthropic or OpenAI as your AI provider with a curated model list, alongside Ollama and LM Studio. Everything local keeps working exactly as before, and the app asks before anything leaves your device. — all changes

Most AI note tools ask you to upload your thinking to somebody else's servers. This one doesn't ask, because it has nowhere to send it.

What "private" means here — specifically

• No account. There is no sign-up, because there is no server to sign up to.
• Local AI. The app installs and starts Ollama or LM Studio for you and fetches the model you choose. Prompts and map content go to a process on your own machine — or to one server you share across your network.
• Encrypted at rest. The database and settings are AES-256 encrypted (SQLCipher) with your passphrase — so you can keep them on Dropbox, Drive or any storage you don't fully trust.
• On-device search. Semantic search uses an embedding model bundled with the app. Your queries are never transmitted.
• Your folder. Maps and notes live in a directory you pick, and you can back it up or delete it yourself.

The single outbound request the app makes on its own is an update check, which you can switch off. Full detail on the privacy page.

Who this is for

• Researchers and professionals handling material that contractually or ethically cannot be pasted into a cloud chatbot.
• Local-LLM users who already run Ollama or LM Studio and want a real application on top of it, not another wrapper.
• Students on unreliable or metered connections who need study tools that work with the network off.
• Anyone who has looked at a "your data is safe with us" page and wanted a tool that simply doesn't collect the data in the first place.

Being straight with you

MindMap Chat is free to use but not open source (Elastic License 2.0), and the source is not published today. You can read the licence, and the privacy claims above are ones you can verify with a network monitor — but if a fully auditable build is your bar, that is a fair reason to wait.

A local model also won't match a frontier cloud model on raw quality. The trade is deliberate, and it is the default: good-enough AI that never sees your data beats excellent AI that does.

If some maps are not sensitive and you would rather have the frontier model for those, the app can talk to Claude or OpenAI — you choose the provider, you accept a dialog spelling out what gets sent, and you pay them directly. It is never on by default and never chosen for you, and every word on this page still describes what happens when you leave it alone.

Try it

Free, no account. Download — or first check whether your machine can run a local model (it probably can, and the app is useful even without one).