github.com via Hacker News

Inkeep Open-Sources OpenKnowledge, an LLM Wiki and Local Editor

TL;DR

  • OpenKnowledge is a local-first markdown editor with native MCP, skills, and agentic search built for both humans and AI agents.
  • The project hit number one on Product Hunt and Hacker News at v2.0 launch, collecting 1.4k signups in 24 hours.
  • Licensed under GPL-3.0-or-later, the GitHub repository shows 296 stars, 613 commits, and 170 releases at v0.18.0.

OpenKnowledge hit the front page of Hacker News and topped Product Hunt following its v2.0 public release on June 3, drawing 1.4k new signups in the first 24 hours, according to the project's website. The pitch is compact: "A rich text editor for you and your agents. Private, open source, and free."

The design serves two classes of users from the same document store. For humans, it delivers full WYSIWYG editing that the project says makes "markdown files feel like editing a Google Doc or Notion page." For agents and coding harnesses, it ships with built-in MCP (Model Context Protocol) for integration with Claude, Cursor, and Codex, alongside skills and agentic search, framed as enabling "LLM Wikis and agent second brains." All documents stay in plain markdown files backed by git and GitHub, rather than a proprietary database.

The GitHub repository records 296 stars, 613 commits, and 170 releases with v0.18.0 as the latest, and the project is licensed under GPL-3.0-or-later. That license opens the code fully but may constrain teams that need to embed the tool in proprietary commercial products.

What neither the repository nor the marketing site explains is how agentic search operates at the implementation level: whether retrieval runs locally or depends on external model API calls, and what the data story is for teams syncing knowledge through GitHub. For developers already running Claude Code or Cursor workflows who want a structured knowledge base those agents can read and write, OpenKnowledge is a concrete option to evaluate, with the practical value of its agentic features most visible once tested against a real workflow.