Reor

Private, local AI note-taking app that auto-links notes and runs models offline

Hand-reviewed by the Toolglade team·Updated Aug 2026

note-taking#open-source#local-ai#privacy#knowledge-management

Free plan Claimed Self-hosted

Toolglade’s take

Reor is a rare fully-local, AI-native note app, and running models on-device with Ollama is a real privacy win over cloud PKM tools. Auto-linking and RAG-over-notes work well for people with large Markdown corpora. It is early-stage and requires some comfort with local models to get the most out of it. Best for privacy-focused, technically-inclined note-takers who already like tools like Obsidian.

About Reor

Reor is a free, open-source, local AI note-taking app that auto-links related notes, runs RAG and semantic search on-device via Ollama, and keeps all data local.

Reor is a personal knowledge management app built on the premise that AI tools for thought should run locally by default. Notes are stored locally as Markdown and edited in an Obsidian-like editor, while an embedded vector database surfaces related notes automatically in a sidebar as you write. Under the hood, Reor combines Ollama (for local LLMs like Llama 3), Transformers.js, and LanceDB to enable on-device embeddings and question-answering over your corpus. It can generate flashcards, run semantic search, and provide a writing assistant, all without sending data to the cloud. It is completely free and cross-platform across Windows, Mac, and Linux.

TL;DR

Reor is a free, open-source desktop note app that runs AI locally to auto-link notes, answer questions, and search semantically, keeping all data private.

Company overview

Reor is an open-source project (reorproject on GitHub) built around the idea that AI tools for thought should run locally. It is community-driven and free, with no commercial tiers as of 2026.

The project stands on open building blocks—Ollama, Transformers.js, and LanceDB—to deliver on-device AI without cloud dependencies, appealing to the privacy and local-first communities.

Product features

Reor stores notes as local Markdown and edits them in an Obsidian-like editor. Its embedded vector database auto-links related notes, powers semantic search, and enables RAG-based Q&A over the user's corpus.

Additional features include AI flashcard generation and a local writing assistant, all running on-device through Ollama-hosted models, so no note content leaves the machine.

Target market

Reor targets privacy-focused, technically comfortable individuals—researchers, students, and Obsidian-style power users—who want AI features without sending data to the cloud.