# Why I Built Jotaid: Turning Saved Notes Into Connected Knowledge *By Leeron Zhang · 2026-10-10 — Why I built Jotaid, a personal knowledge engine for Mac and iPhone: concept-linked notes, a co-occurrence matrix, themes, and read-only MCP for AI agents.* > **In short** > > **Jotaid is a personal knowledge engine: a native Markdown note-taking app for Mac and iPhone that links concepts, not just notes.** You write ordinary notes and mark the ideas in them with `[[ ]]`. Each idea becomes a concept node with its own page. Jotaid then shows how those concepts relate: a knowledge graph, a co-occurrence matrix that shows how strongly two ideas belong together, and a Trend view of how they change over time. Notes grow into themes and projects on a visual canvas. Optional AI proposes concepts and themes, but nothing is created until you accept it. On the Mac, AI agents can read your notes through a read-only MCP server. The free version is a complete single-space note app, not a trial. There is a question I have never quite been able to put down: when we use a note-taking app, are we saving information, or building knowledge? Over the years I have written down a great many things. Stray ideas, reading highlights, research material, work notes. Each one seemed worth keeping at the moment I wrote it. But as the pile grew, I noticed something uncomfortable: **information you write down does not turn into your own knowledge on its own.** You can sort notes into folders, organize them with tags, search them by keyword, connect them with backlinks. All of these rest on the same assumption: that you already know what you are recording, what you will look for later, and which notes belong together. The connections that matter most are usually the ones you did not see coming. The happy accident. The link you had not noticed was there. That is why I started building Jotaid. I wanted a tool that does more than hold what I write. I wanted one that would help me slowly see the shape of what I know. ## Why more notes don't add up to more knowledge Note-taking tools, and the personal knowledge management (PKM) apps built on them, have mostly solved two problems: capturing and finding. Over time the ways of organizing have grown richer, from folders and tags to backlinks and knowledge graphs. But they share one trait: **most of the structure is yours to build and maintain.** Here is an example. I had a few notes about *artificial intelligence*, *personal knowledge management* and *cognitive load*. I wrote them at different times, in different projects. When I wrote them, I did not think they had anything to do with each other. Only later, after more reading and more thinking, did I see that they all pointed at one question: how do our tools change the way we think? If I had only been using folders, those notes would have stayed where they were, each in its own corner. Even with backlinks, I would have needed to realize, first, which notes to link. Knowledge does not grow in folders. A concept turns up again and again in different situations. One question touches several fields at once. New understanding often comes from pieces that were scattered being put back together in a new way. I wanted a note-taking tool that would help me watch that happen, not just store each thing I wrote. Call it a second brain if you like, but I care less about how much it remembers than about whether it helps me see what my notes add up to. ## From backlinks between notes to links between concepts This is where Jotaid differs most from the note apps I used before. You still write notes the ordinary way. But on top of the notes there is a second layer: concepts. In Jotaid each concept is a **node**. In most note apps, a backlink connects one note to another. In Jotaid, the question I care about is a different one: **which concepts do different notes share, and how are those concepts related?** Say one note about knowledge management mentions *memory*, *cognitive load* and *artificial intelligence*. Another note, about how to study, mentions *memory* and *cognitive load* again. While writing, I put `[[ ]]` around *memory*, and it becomes a node. I don't have to create it somewhere else first. It exists the moment I write it. When I reread an old note, I can also select a word and mark it as a node. If you practice Zettelkasten, this will feel familiar: nodes play roughly the role of the keyword index in Niklas Luhmann's slip box, a set of entry points into a web of notes. ([Zettelkasten in Jotaid](https://docs.jotaid.com/en/guide/zettelkasten) walks through the workflow.)  *The graph focused on Memory. The panel on the right lists the 19 notes that link to it, then the concepts it most often appears with, from Forgetting Curve down to Knowledge Graph.* So *memory* is no longer just a word that happens to appear in two notes. It is an object in its own right. A node has a body of its own, where I can write a definition, my current understanding, and where it ends and its neighbors begin. When I open it, I can see which notes mention it and which concepts it tends to appear with.  *Memory as a page of its own. The definition and the five points under it are written on the concept itself, not in any one note, and they link onward to neighboring concepts like Cognitive Load and Spaced Repetition.* That means I don't need to remember the title of some old note. I can start from a concept and walk back into what I wrote before. And as the same concepts keep appearing across different notes, scattered pieces have a chance to join into a web that gets clearer over time. That is the core idea behind Jotaid: **notes carry knowledge, but concepts and the relationships between them are how you come to understand it.** (I make the longer case for this in [Why Backlinks Are Not the Same as Relationships](https://jotaid.com/blog/posts/backlinks-are-not-relationships).) ## Beyond the knowledge graph: seeing how strongly ideas connect Many note apps now have a graph view, and I like it too. It takes the connections buried in a list and makes them visible at a glance. But after using one for a while, I found that once there are enough nodes, a screen full of dots and lines looks impressive without helping me understand much. So in Jotaid I tried to go one step further: not only to show *whether* two ideas are connected, but *how closely*. Alongside the graph, Jotaid has a **co-occurrence matrix**, a grid for seeing which concepts appear in the same notes. Every concept gets a row and a column. If *cognitive load* and *personal knowledge management* keep turning up in the same notes, the cell where they cross gets darker. The more notes two concepts share, the stronger their relationship reads.  *The darker the cell, the more notes two concepts share. The highlighted cell pairs Personal Knowledge Management with Cognitive Load: seven shared notes, listed in the panel, where the AI's reading ends on a question rather than a verdict.* There is also a **Trend** view, which shows when a concept first appeared, when it grew, and when it went quiet.  *One row per concept. Each line starts where the concept first appeared, spikes where it was busy, and runs flat where it went quiet. The sentence at the top sums up how the project's focus moved.* Together they show me which questions I keep coming back to, which concepts tend to travel together, and whether things that looked unrelated are starting to connect. Of course, **appearing together is not the same as causing each other**, and it does not mean two concepts are related in any formal or academic sense. For me, these views are prompts: they point at what is worth rereading, comparing and thinking through again. The actual understanding is still yours to do. (If you want to read your own matrix more closely, I wrote a guide: [What a Co-occurrence Matrix Can Reveal About Your Notes](https://jotaid.com/blog/posts/co-occurrence-matrix-notes).) ## From scattered notes to themes: design thinking for personal knowledge management Beyond relationships between concepts, there was another problem I wanted to solve: how does a pile of scattered notes slowly become a coherent topic? Is there a way to let notes organize themselves as they grow? With other note apps I kept going back and forth between two approaches. One is to build a strict system of categories before writing anything, and put every note in its place. The other is to write everything down first and sort it out once there is enough. The first makes writing heavy. The second makes the pile grow. So Jotaid has **themes**. A theme is not just a folder. It is a way of gathering notes around a direction of research, a question, or an idea.  *The theme Personal Knowledge Tools in the AI Era on the canvas: 31 notes as cards inside one border. The next theme, Learning and Memory, begins underneath.* For example, I had been writing notes about *AI agents*, *personal knowledge management* and *backlinks*, and eventually saw that they all belonged to something bigger: personal knowledge tools in the age of AI. At that point I can create a theme and gather those notes into it. On the canvas, every note is a card I can move. I can arrange, group and rearrange them the way you would sticky notes on a desk. And the same theme can be seen as a list of notes, or from the angle of the concepts it contains. Jotaid's three tiers, **project, theme and note**, follow the "diverge, then converge" rhythm of design thinking. Notes are where you keep diverging. A theme is the first convergence. A project is where everything finally comes together. To me this is closer to how thinking actually works. Most of the time we don't start with a clear framework and then fill it in. We record, compare and rearrange, and only through that do we slowly find out what we are actually thinking. ## AI-assisted note-taking that leaves the judgment to you While building Jotaid, I kept asking what role AI should play. Having AI summarize an article, generate notes or answer questions is nothing new anymore. But I did not want AI to be a chat window bolted onto a notes app. Rather than having it write for me, I wanted it to take on the repetitive work that machines are actually suited to. Once notes have piled up, for example, **Smart Node** reads through a project's notes, proposes concepts that deserve to be nodes, and lists the notes that mention each one. **Smart Theme** proposes themes and the notes that belong in each. What they do is close to affinity mapping in design thinking. The theme names Smart Theme suggests tend not to be category words like "user research", but short phrases that name a tension.  *Smart Node found 35 candidate concepts, each with a one-line definition and a relevance rating. Only the ticked one, Empathize, will be created when you press Apply.* AI can also summarize a theme, in four parts: an overview, key findings, insights, and "How Might We" questions. It can also explain how two concepts might be related. Behind all of these I held to one principle: **AI can make suggestions. Organizing and judging stay with you.** Smart Node and Smart Theme results only take effect when you tick them one by one. Anything you leave unticked is never created. A concept or a theme should not become part of your knowledge structure just because a model found it plausible. Personal knowledge management is not only about information. It is also about the way you, the person using it, make sense of things. (This is the subject of a whole essay I want to write: AI should help you notice connections, not decide what they mean.) AI features in Jotaid run on your own API key from an AI provider. Which provider your notes go to, and which model reads them, is your decision. ## Your notes as context for AI agents (MCP) As AI agents develop, I have started thinking about another question. If AI is no longer just something you chat with, but a partner that keeps helping you get work done, can everything you have written over the years actually be put to use? Today we keep explaining our background to AI again and again: the state of a project, the ideas we already have. Or we copy long passages out of our notes and paste them in as context. Very often, all of that is already written down in our own notes. So Jotaid on the Mac provides an **MCP** (Model Context Protocol) server. AI agents that support the protocol can, within the projects you allow, browse your projects and notes, search them by meaning or by keyword, look up which notes mention a concept, see which concepts keep appearing together and how a concept has changed over time, and read your concept network.  *Settings → Agent. The switch runs a read-only server on this Mac only (note the 127.0.0.1 address), and Readable projects decides what an agent is allowed to see.* That means when I talk a project over with an agent, it is not only reading the text of one note. It is reading the concept and theme structure Jotaid has already built: the same web I see in the graph, matrix and Trend views. A few boundaries, stated plainly: - It is **read-only**. Agents cannot create, edit or delete anything through it. - It is **off by default**. You turn it on yourself in Settings, and choose which projects agents may read. - The server listens only on your own Mac. Other devices on your network, and the internet, cannot reach it. - It is part of Pro. This is only a beginning. But I have come to believe that the value of a personal knowledge tool may not be limited to helping you look back. It may also be in turning what you have built up over years into personal context that your AI tools can keep using. (I wrote more about this direction in [What Is a Personal Knowledge Engine?](https://jotaid.com/blog/posts/personal-knowledge-engine)) ## A native Markdown note app first ![A note open in Jotaid's Markdown editor on the Mac, containing a LaTeX formula, a table of concept pairs, two [[wiki links]] and a highlighted phrase](https://jotaid.com/blog/media/why-jotaid-markdown-editor.webp) *A note in the editor: a LaTeX formula, a table, two [[links]] to concept nodes and a highlighted phrase, all written in plain Markdown.* I have talked a lot about concepts, relationships and AI, but Jotaid is a note-taking app first. You write in Markdown, with live syntax highlighting, tables, code blocks and LaTeX math, and with backlinks. On the Mac, while Jotaid is running, ⇧⌘J brings up Quick Note from any app so you can jot something down. On the iPhone you can capture an idea whenever it comes. Both apps are native, and they sync through iCloud. Jotaid is local-first: it works fully offline, and iCloud sync is an optional layer on top. It imports and exports Markdown. Import a folder, and its structure becomes a project and its themes. If you are coming from a folder-based Markdown app such as Obsidian or Typora, a vault or folder usually comes in one step, with the original file timestamps kept. Export, and each note is an `.md` file with YAML front matter, its images in an `assets` folder next to it. Easy to move in, and just as easy to move out. I don't want anyone to have to learn a whole methodology before they can build a knowledge structure. Ideally you just write as usual, organize when you feel like it, and go into the concept and theme views when you want to dig deeper. Not stop before every note to decide which folder it goes in, which tags it gets, and how many links to make. So even if you never touch the AI, you can use Jotaid as an ordinary note-taking app and build your own structure by hand. That part is complete in the free version. ## Why build another note-taking app? I asked myself this many times while building it. There are already plenty of mature note apps. Some are great for quick capture, some for long-form writing, some offer flexible databases, some have rich plugin ecosystems. I don't think everyone needs to replace what they use. If you mostly keep a journal, save reference material or jot down the occasional idea, many existing tools are already good enough. Jotaid is for a different need. When you read, research and write over a long time, or keep building up material around a few questions, simply saving things stops being enough. You start to want to know: what have I thought about before? Which questions keep coming back? Do these different-looking pieces connect? Can all of this slowly add up to a fuller understanding? I want Jotaid to offer a way of looking at those questions that is different from managing files. Adding AI does not mean it will do the understanding and thinking for you. It cannot automatically turn every note into knowledge. What it can do is take some of the repetitive organizing off your hands, and make the connections scattered across different times and different notes easier to see. ## Frequently asked questions ### What is Jotaid? Jotaid is a personal knowledge engine for Mac and iPhone: a native, local-first Markdown note-taking app that treats the concepts in your notes as first-class objects. It links concepts rather than only notes, shows their relationships in a knowledge graph, a co-occurrence matrix and a Trend view, and lets notes grow into themes and projects on a visual canvas. ### How is a concept node different from a backlink? A backlink connects two notes. A node is a concept, such as *memory*, with its own page: its own written body, the list of notes that mention it, and the concepts it most often appears with. You create one by wrapping a word in `[[ ]]` while you write, or by selecting a word in an existing note and marking it as a node. ### What does the co-occurrence matrix show? It is a grid of your concepts, one row and one column each. The cell where two concepts cross gets darker the more notes they share, so you can compare how strongly any pair belongs together across the whole project. Appearing together is a prompt to look closer, not proof that two ideas are causally or formally related. ### Do I have to use AI? No. Everything that analyzes the links you make yourself, including the graph and the matrix, works without AI and is free. AI is optional, part of Pro, and runs on your own API key from the AI provider you choose. Its proposals, such as Smart Node concepts and Smart Theme groupings, only take effect when you accept them. ### Can AI agents read my Jotaid notes? Yes, on the Mac, through a Model Context Protocol (MCP) server. It is read-only, off by default, limited to the projects you allow, and listens only on your own Mac. Any agent that supports MCP servers over HTTP with a custom header can connect. It is part of Pro. See [Connect AI Agents (MCP)](https://docs.jotaid.com/en/guide/mcp). ### Can I bring my Markdown notes in, and take them out again? Yes. You can import Markdown files or whole folders, with folders mapped onto projects and themes. You can export notes, or whole projects on the Mac, as `.md` files with YAML front matter and their images. See [Import & Export](https://docs.jotaid.com/en/guide/import-export). ### Is Jotaid free? Jotaid is free to download, and the free version is not a trial: the Inbox, a single space, is fully featured. Pro adds multiple projects, AI, bulk move, cross-filtering and MCP on the Mac, through a monthly or yearly subscription or a one-time purchase that all unlock the same features. ## Where it stands Jotaid is on the App Store for Mac and iPhone. The free version is not a trial. Inside the Inbox, a single space, you get the full editor, backlinks, the knowledge graph, the co-occurrence matrix, the canvas, iCloud sync and on-device related notes, for as long as you like. Pro adds multiple projects, the AI features (Smart Node, Smart Theme, theme summaries and more), bulk move, filtering by several conditions at once, and agent connections (MCP) on the Mac. You can unlock Pro with a monthly or yearly subscription or a one-time purchase. All three unlock the same features. ([Free vs Pro](https://docs.jotaid.com/en/guide/pricing) has the full list.) It is still evolving. As an independent developer, I have many ideas I have not built yet, and plenty of design decisions that need to change based on what real users tell me. But one direction has not changed through the whole process: **I want notes not just to be kept, but to keep forming connections and surfacing themes, so that what you write over the years slowly becomes worth more.** If you have also built up a large pile of notes, and wondered how to turn them into knowledge you can actually use, I would love to hear how you work today. Do you rely on folders, tags, backlinks, or a system of your own? And if you try Jotaid, tell me what fits the way you work and what still falls short. You can reach me at contact (at) jotaid (dot) com.