OpenKnowledge: An AI-First Open Source Alternative to Obsidian and Notion

OpenKnowledge is an open-source, "what you see is what you get" (WYSIWYG) markdown editor designed to bridge the gap between the local-first flexibility of Obsidian and the collaborative power of Notion. By integrating directly with AI agents and leveraging Git for versioning and synchronization, it provides a collaborative environment for technical teams to build AI-enhanced knowledge bases.

Core Features and AI Integration

OpenKnowledge focuses on a seamless side-by-side experience between human writing and AI generation. Unlike traditional markdown editors that rely on community plugins for AI, OpenKnowledge provides native integrations with several major AI tools:

  • Direct Agent Integration: The tool integrates with Claude, Codex, and Cursor desktop apps. AI agents can open an OpenKnowledge editor within their embedded web browsers, allowing for a side-by-side workflow where the agent and human edit the same document.
  • AI-First Infrastructure: The platform includes built-in Model Context Protocol (MCP) servers, skills, and Retrieval-Augmented Generation (RAG) capabilities specifically for "AI Second Brain" scenarios and specification writing.
  • Hybrid Interface: Users can access the tool via a macOS desktop application featuring a file navigator and link explorer, or through a CLI and embedded terminal for users who prefer a TUI-first (Text User Interface) approach.

Technical Architecture

The development of OpenKnowledge involved solving several complex synchronization and fidelity challenges to ensure that markdown files remain the source of truth while providing a modern editing experience.

Bidirectional Markdown Pipeline

One of the primary engineering hurdles was creating a pipeline to convert ProseMirror (which uses Abstract Syntax Trees or ASTs) to markdown in a bidirectional, lossless way. Because ASTs are not designed for byte-fidelity, the team developed a custom pipeline to ensure that the markdown files on disk remain exactly as intended.

Dual-Observer CRDT

To maintain synchronization between the ProseMirror editor state and the markdown file state, OpenKnowledge utilizes a dual-observer Conflict-free Replicated Data Type (CRDT) system. This architecture, combined with Git, enables:

  • Collaborative Editing: Real-time visibility into what AI agents are modifying in the markdown.
  • Version Control: Full undo/redo capabilities and comprehensive version history.
  • Private Team Sync: "Share" and cloud-sync functionality that leverages GitHub under the hood, ensuring data remains private and version-controlled.

The Tech Stack

The project is built using a modern web-technology stack wrapped in Electron for the macOS app:

  • Editor Frameworks: Tiptap/ProseMirror and CodeMirror.
  • Synchronization: yjs (CRDT).
  • Search and Parsing: Orama, remark, rehype, micromark, and mdast.
  • Platform: Electron (macOS).

Community Feedback and Considerations

Following its launch on Hacker News, the community highlighted several strengths and potential limitations of the current implementation.

Platform and Model Constraints

Users noted that the current macOS-only availability is a significant limitation for those on Windows, Linux, or Android. Additionally, there is a strong demand for support for local LLMs (e.g., via LMStudio or OpenAI-compatible endpoints) to ensure a fully private, local-first experience, as some users argued that calling the tool "open source" while tying it to proprietary AI services is a contradiction.

Comparison with Existing Tools

Community members compared OpenKnowledge to Obsidian and Notion, noting that the primary gap in the current market is a tool that combines Obsidian's local markdown files (ideal for LLMs) with Notion's multiplayer collaboration features.

"Obsidian: great for LLMs (local markdown files), bad for collaboration (no multiplayer features like multi editor, comments). Notion: not great for LLMs (network round trips, block-based editing), great for collaboration."

UX and Workflow Concerns

Some users expressed a preference for the "not quite WYSIWYG" nature of Obsidian, where markdown syntax is visible during typing. Others raised concerns about the Electron-based UI, suggesting that a native app would provide a more polished experience in terms of scrolling and selection. There were also reports of a bug where the installation process may interfere with Codex config.toml files.

Summary of Capabilities

Feature OpenKnowledge Obsidian Notion
Format Local Markdown Local Markdown Proprietary Blocks
UI WYSIWYG Hybrid/Live Preview WYSIWYG
AI Integration Native (Claude/Codex/Cursor) Plugin-based Built-in
Sync/Collab Git/GitHub (OSS) Proprietary Sync/Git Plugins Cloud-native
Platform macOS Cross-platform Cross-platform

Sources

Related

  • Project
    inkeep/open-knowledge

    A private, local WYSIWYG markdown editor with integrated AI agent support for building knowledge bases and LLM wikis.

  • Project
    AgriciDaniel/claude-obsidian

    A local-first knowledge system for Claude Code and Agent Skills hosts that turns source material into linked, source-cited Obsidian pages with strict provenance tracking.

  • Project
    eugeniughelbur/obsidian-second-brain

    An AI-powered second brain for Obsidian that allows LLMs to search, rewrite, and maintain a persistent knowledge vault across different AI sessions and platforms.

  • Dispatch
    Integrating AI with Obsidian: Why Generative Content Can Degrade a Second Brain

    Over-reliance on AI-generated content within a personal knowledge base like Obsidian can diminish cognitive ownership and dilute the value of a 'second brain' by replacing active thinking with synthetic noise.

  • Project
    nashsu/llm_wiki

    A desktop application that automatically transforms documents into a structured, interlinked personal wiki using LLMs, featuring a knowledge graph and automated deep research.