Study with an AI mind map
A chatbot gives you a long answer to one question. A mind map gives you the shape of a whole subject. MindMap Chat puts the two together: you build the map, and you ask the AI about one branch at a time.
Why a map helps you learn
Most of understanding a subject is knowing how its parts relate: what depends on what, which idea is an example of which. A map shows that at a glance, and building it makes you decide it. A chat transcript does the opposite. After an hour it is four hundred messages long, the good answer is somewhere in the middle, and nothing in it tells you how the ideas connect.
In MindMap Chat the conversation hangs off the map. Each node can have its own chat, and the AI answering it sees that node, its notes, its parent and its children, plus an outline of the map around it. Ask "why does this follow from its parent?" and it knows what the parent is.
A way to work through a topic
- Start from a template. In the library, ⋮ → Start from a template offers a ready structure for studying, for reading notes and for a decision. Or start a blank map and write the topic in the central node.
- Sketch the outline yourself. Add the chapters, lectures or exam areas you already know about as branches. Getting this first draft down is where the learning starts.
- Let the AI fill gaps. Use Extend on a branch and the AI proposes new nodes under it. Read them before you keep them: Apply adds them as one step, and one Undo removes the lot.
- Ask about one node at a time. Open Chat about node with AI on whatever you do not understand yet. When an answer is good, save it as that node's note, so it stays where it belongs instead of in a transcript.
- Research what needs depth. AI Research, under a node's properties, answers a question straight into the node's note.
- Step back. Collapse the branches you know to see the shape of the whole, and open a map chat to ask how the parts fit together. A map chat never edits the map.
Finding things again
A study map grows. Switch the search to the all-minilm-l6-v2 model in
Settings → AI and the search bar finds nodes by meaning, not only by the exact
word: search for "cell energy" and it can find the node you titled "mitochondria". (The lighter
default model matches only words the node and the query share.) Both models are built into the
app, so search works on a train with the network off. It reads a node's title first, so a title
that says what the node is about makes it easier to find.
Notes that hold real material
- Rich text in notes and details: select text to format it.
- Maths written in LaTeX renders as formulas.
- Links in a note open in your browser.
- Icons, colours and shapes to mark what you have mastered and what you have not.
- Papers from arXiv. Find papers on a node lists research papers on its topic from a public dataset of arXiv papers, and the AI's answers can draw on them. It uses public data only, and one switch turns it off.
An AI that is yours
The AI runs on your own computer, through Ollama or LM Studio, so there is no subscription and no per-question fee, and your notes are not sent to anyone. The app can install the runtime and download a model for you; a model of about 4 to 5 GB runs on a laptop with 8 GB of memory. If you prefer a cloud model, Claude and OpenAI work with your own API key, after you agree.
Your maps are stored on your device, encrypted with your passphrase, with no account. Every edit, including every change the AI makes, can be undone.
Already have maps?
If your notes are in Freeplane .mm or XMind files, import them and carry on: every
import is a new copy, and your original file is left as it was. See
opening Freeplane files.