REF / SOFTWARE

My Notebook — AI- and MCP-Native Personal Notes App

A personal notes app with AI integration, native MCP support, and bespoke UIs for 11+ note types — from YouTube and reading lists to tasks and expenses.

RoleSolo Developer
Year2026
Outcome11+ custom note types, AI + MCP-native
DomainSoftware
00
STACK

Tech used.

Next.jsTypeScriptGemini GenAI APIMCPFirebase

The Problem

Every notes app I tried made the same mistake: it flattened everything into one text box. A YouTube video I want to watch later, a book I'm halfway through, a bank account, a login I need to remember, this week's tasks, a running list of expenses — they all got dumped into the same undifferentiated markdown blob and lost.

But those things aren't the same shape. A watchlist wants a poster and a "seen it?" toggle. A reading list wants an author and a progress state. A credential wants a masked field you can copy, not plaintext in a paragraph. A bank account is structured data, not prose. Treating them all as "a note" is the reason personal knowledge tools quietly rot.

So I built my own — a notes app where the type of a note determines its interface, and where an AI can actually read and act on the whole thing.

What I Built

My Notebook is organized around three pillars.

1. Typed notes with bespoke UIs. Creating a note starts by choosing what it is. Each type gets a purpose-built editor and view rather than a generic text field. The types that ship today:

Note typeWhat it captures
Text noteFree-form writing, the classic default
YouTubeVideos to watch, saved with their context
Reading listArticles and long-reads queued up
MusicTracks and albums to come back to
WatchlistFilms and shows to watch
BooksReading, with author and progress
Bank accountsStructured account details
CredentialsLogins, stored as structured fields
TasksTo-dos with state
ExpensesMoney in and out
ContactsPeople, as records rather than prose

2. AI integration. The notebook isn't a passive archive — an AI layer can summarize, search across, and answer questions about everything in it, so the notes become something you can actually query instead of a folder you're afraid to open.

3. MCP support. The whole notebook is exposed over the Model Context Protocol, so external AI clients like Claude can connect to it as a set of tools and read or write notes directly. That turns the app from a place I type things into a place my AI assistants can operate on my behalf.

11+
Custom note types
each with a bespoke UI
AI
Native, not bolted on
summarize + query across notes
MCP
Server built in
Claude and other clients as tools
1
Daily driver
the tool I actually use

Why I Built It

This one started as a tool purely for myself. I wanted a single place that respected the difference between a task and a tweet and a bank login, and that my AI assistants could plug into rather than sit outside of. It became my daily driver — the notebook I reach for first — and a small proving ground for the same idea I keep coming back to in my other work: the interface should bend to the data, and AI should be a first-class citizen of the app, not a chat window bolted onto the side.