open-knowledge

OpenKnowledge is a markdown‑based editor that presents a visual, Notion‑style canvas while keeping the underlying files as plain markdown. The application runs locally on any operating system and stores every document as a regular .md file, so you retain full ownership of the data. It is marketed as “AI‑native,” meaning that large‑language‑model agents can be plugged in to read from and write to the same files you edit.

The editor supplies a collection of inline components that you add through the toolbar or keyboard shortcuts. A callout block creates a highlighted note with optional icons; an accordion lets you hide and reveal sections with a clickable summary; tabs let you switch between multiple panes of content without leaving the page. Mermaid syntax is rendered directly into flowcharts and diagrams, while media blocks let you drop in images, resize them, add captions, or embed videos with playback controls. Code snippets are displayed with syntax highlighting, and tables can be built using standard markdown table syntax. All of these elements are stored as markdown extensions, so the file remains readable outside the editor.

OpenKnowledge includes ready‑made templates for common knowledge‑base items such as launch week recaps, incident postmortems, deployment runbooks, and on‑call guides. When you select a template, placeholder sections appear with headings and sample markdown that you replace with your own text. The editor tracks changes in a git repository behind the scenes, so each edit creates a commit and you can view history or revert to earlier versions. Sharing a document or an entire workspace is a single click; the tool generates a git remote link that teammates or agents can clone, preserving authorship and version control.

Agent integration is provided through a “native MCP” layer that connects to many LLM services, including Claude, Cursor, Codex, Gemini, Copilot, and a long list of community bots. Once an agent is linked, you can issue prompts like “Help me write up our launch week” and the agent will generate markdown content directly in the editor. Agents also have built‑in skills for navigating the knowledge base, adding or updating sections, and performing “agentic search” that uses embeddings and hierarchical retrieval‑augmented generation to locate relevant pages.