OpenKnowledge: open source AI editor takes on Obsidian

Key Takeaways

• OpenKnowledge is a new open source markdown editor with native AI integrations for Claude, Codex, and Cursor
• The tool runs locally, stores data on your machine, and uses Git for collaboration rather than proprietary sync
• MCP (Model Context Protocol) support lets any AI agent access your knowledge base directly

A new open source project called OpenKnowledge launched on Hacker News this week, positioning itself as an AI-native alternative to Obsidian and Notion. The tool combines a WYSIWYG markdown editor with built-in integrations for Claude, OpenAI Codex, and Cursor, all while keeping your data local and free.

The pitch is straightforward: existing knowledge management tools either lock your data in the cloud or treat AI as an afterthought. OpenKnowledge tries to solve both problems at once. It runs on your machine, syncs via Git, and treats AI agents as first-class citizens through MCP (Model Context Protocol) support.

What does OpenKnowledge actually do?

At its core, OpenKnowledge is a markdown editor. It opens any folder containing markdown or MDX files, including existing Obsidian vaults, codebases, or wikis. The interface includes a file navigator, search, tabs, and a graph view for wiki links. Think of it as Notion's editing experience layered on top of Obsidian's local-first architecture.

The editor claims "full true WYSIWYG," meaning you edit formatted text rather than raw markdown syntax. Tables, headers, and rich components render inline as you type. This mirrors Notion's approach but stores everything as plain text files you control.

The macOS app is the primary offering. Linux, Windows, and Intel Mac users can run the same editor as a local web app through npm. Installation requires Node.js 24 or later, then a few commands to initialize a project and launch the browser-based interface.

How does the AI integration work?

OpenKnowledge ships with what the team calls "collaborative AI-editing" for Claude, Codex, and Cursor. The app auto-detects AI tools on your system and configures MCP and skill files accordingly. This lets AI agents read, search, and modify your knowledge base directly.

MCP support is the technical differentiator here. Model Context Protocol, developed by Anthropic, provides a standardized way for AI assistants to interact with external data sources. Rather than copy-pasting context into a chat window, an MCP-enabled agent can query your notes, wikis, and specs programmatically.

The project mentions "agentic search" and "agent second brains" as target use cases. The idea: your personal knowledge base becomes the memory layer for AI assistants. Instead of re-explaining context every conversation, the agent references your existing documentation.

Privacy and collaboration without the cloud

OpenKnowledge stores everything locally by default. There's no account, no telemetry mentioned, and no sync server to trust. Team collaboration works through Git and GitHub. Enable it and changes flow through version control rather than a proprietary backend.

This appeals to developers already comfortable with Git workflows. It also sidesteps the trust problem with cloud-based AI tools. Your documents stay on your machine. The AI integration happens locally through the desktop apps rather than sending files to a remote API.

The GPL-3.0 license means the code is genuinely open source. You can fork it, modify it, self-host it. This contrasts with Obsidian, which is free to use but not open source, and Notion, which is neither.

Where does it fit against Obsidian and Notion?

Obsidian users might find OpenKnowledge interesting for the AI-native approach, since Obsidian relies on third-party plugins for AI features. Notion users who want local storage and open formats might consider the switch, though they'd lose Notion's real-time multiplayer editing.

What's the catch?

This is a brand new project. The GitHub repo is fresh, community support is limited to a Discord server, and there's no track record to evaluate. The macOS app is the polished version. Everyone else gets a web app running through npm.

The AI integrations require you to already have Claude, Codex, or Cursor installed and configured. OpenKnowledge connects to them but doesn't bundle them. You need your own API keys and subscriptions for the underlying AI services.

Frequently Asked Questions

Is OpenKnowledge really free?

Yes. The core app is free and open source under GPL-3.0. However, the AI integrations require separate subscriptions to Claude, OpenAI, or Cursor.

Can I import my Obsidian vault into OpenKnowledge?

Yes. OpenKnowledge can open any folder with markdown files, including existing Obsidian vaults, without conversion.

Does OpenKnowledge work offline?

The editor runs locally and works offline. AI features require internet access to communicate with the respective AI services.

Is there a mobile app for OpenKnowledge?

Not currently. The project offers a macOS desktop app and a web-based CLI version. Mobile support isn't mentioned in the current release.

Source: Hacker News: Best