KnowNote

Overview

KnowNote is a desktop application that turns PDFs, Word documents, PowerPoint files and web pages into a knowledge base you can question. Parsing, splitting, embedding and vector search run inside the app, so retrieval works offline. Answers keep references to the source passages, including where each passage sat in the extracted text.

Key Features

• Local retrieval runs in-process and keeps working offline; documents, notebooks and embeddings stay on the machine.
• Traceable answers keep the source passages and record where each passage sat in the extracted text.
• Bring your own chat model through OpenAI-, Anthropic- or Google-compatible endpoints or a local server such as Ollama.
• Import PDF, Word, PowerPoint and web pages, with each format parsed to preserve page, heading and slide structure.
• Save structured notes beside sources and turn a notebook into a mind map.
• Serve a notebook read-only over stdio MCP so local agents can search documents with provenance.

Use Cases

• Research: question a set of papers and inspect the cited passages behind each answer.
• Study: build notes from course readers, slides and other source documents.
• Agent-assisted reading: let a local MCP client search an existing library instead of pasting documents.

Getting Started

• Download the Windows setup executable or the macOS Apple Silicon disk image from GitHub Releases.
• On macOS, open the disk image, drag KnowNote into Applications, and run sudo xattr -rd com.apple.quarantine /Applications/KnowNote.app once before launching.
• On Windows, if SmartScreen appears, choose More info and then Run anyway.
• On first run, open Settings, add at least one model connection under Models, then start asking questions.

Deployment & Requirements

• Desktop app for Windows x64 and macOS Apple Silicon; no Linux build is available.
• No Docker, no server and no account are required.
• At least one model connection must be configured; it can be any OpenAI-, Anthropic- or Google-compatible endpoint or a local server such as Ollama.
• The local embedder downloads once and then runs on the machine, while notebooks, notes and embeddings stay local.
• The macOS build is unsigned and requires a one-time quarantine-clearing step; Windows may show a SmartScreen prompt.

Before You Adopt

• License: GPL-3.0. Review its terms before using, modifying, or distributing the project.
• Unsigned, non-notarised builds trigger first-launch security prompts on macOS and Windows.
• There is no Linux build, and macOS support is limited to Apple Silicon.
• When asking a question, only the retrieved context for that request is sent to the configured model endpoint.