# OpenKnowledge

> A local-first markdown IDE for knowledge bases humans and agents both edit

- Category: AI Productivity & Writing / Notes & Knowledge Bases
- Language: TypeScript
- Licence: GPL-3.0-or-later
- Repository: https://github.com/inkeep/open-knowledge
- Website: https://openknowledge.ai
- Canonical page: https://ai-tldr.dev/tools/openknowledge/
- Updated: 2026-06-25

## In short

OpenKnowledge is a markdown editor built for knowledge bases that both people and AI agents read and write. It sits in the Notes & Knowledge Bases category of the AI Productivity & Writing tools and is open-source, written in TypeScript and licensed GPL-3.0-or-later. To get started: install the web app.

## Overview

OpenKnowledge is a markdown editor built for knowledge bases that both people and AI agents read and write. Inkeep describes the goal as the editing feel of a document app with the file handling of a code editor: your content stays as plain markdown files on disk, but you get true WYSIWYG editing, tabs, search, a file tree and a visualisation of the wiki links between documents. Because the store is just a folder of markdown, a repository of notes can be synced and reviewed with git and GitHub instead of a proprietary database.

The AI side is deliberately side-by-side rather than baked in. OpenKnowledge runs alongside CLI coding harnesses — Claude, Codex, OpenCode and Pi — and exposes the knowledge base through MCP and agent skills, so an agent can search the wiki as a graph and write back into it. That is what the codebase-wiki starter pack shipped in v0.18.0 does: point it at a repository and an agent drafts a wiki describing the code, which you then edit by hand like any other document.

It ships both as a desktop app for macOS, Windows and Linux and as a web app you install from npm and run against a local folder, so the same knowledge base can be opened on a laptop or served to a team. The project is licensed GPL-3.0-or-later and written in TypeScript.

## What it does

- True WYSIWYG markdown editing over plain files on disk — no proprietary store
- Desktop and web apps with a file tree, tabs, search and wiki-link visualisation
- Side-by-side integration with Claude, Codex, OpenCode and Pi through CLI harnesses
- MCP server and agent skills that let an agent search the knowledge graph and write documents
- Git and GitHub sync for sharing and reviewing a knowledge base without a backend
- Embeddable HTML components for specs and reports inside a markdown page

## Getting started

Install the desktop build if you want an app, or the npm package if you want to run it over an existing folder. The web app needs Node.js 24+ and git.

1. Install the web app

   The npm package installs the `ok` command globally.

```bash
npm install -g @inkeep/open-knowledge
```

2. Initialise a knowledge base and open it

   Run this inside the folder you want to edit — an existing repo of markdown works as-is.

```bash
cd your-project
ok init
ok start --open
```

3. Or install the desktop build

   Downloads for macOS, Windows 10+ (x64/Arm64) and Linux (deb/rpm, x64/arm64) are published at openknowledge.ai/download. Once a base is open, point a coding harness at it and use the MCP server and skills to let the agent search and write documents.

```text
https://openknowledge.ai/download
```

## When to use it

- Keep a team wiki as plain markdown in git so changes are reviewed like code
- Generate a first-draft wiki of an unfamiliar codebase, then edit it by hand
- Write specs and reports that an agent can later search through MCP
- Work through notes locally when the content should not leave the machine

## Version history

- 2026-06-25 v0.18.0: v0.18.0 added a codebase-wiki starter pack that scaffolds an agent-written wiki of a repository, alongside side-by-side AI editing with Claude, Codex and Cursor in the local-first markdown app.