Claude Code

Open Knowledge plugs into Claude Code through MCP plus an installed skill that teaches the agent how to use the tools.

Install

The Open Knowledge desktop app handles wiring. On first launch (or any time you open a project for the first time), the consent dialog detects Claude Code on your machine, lists it as one of the editors it can configure, and writes the MCP server entry plus the project skill for you.

It registers the MCP server in Claude Code's user config (~/.claude.json) and project config (.mcp.json), and installs a skill at .claude/skills/open-knowledge/SKILL.md. Re-trigger the dialog by deleting ~/.ok/mcp-status.json and relaunching.

Verify

Open the project in Claude Code and ask:

List all the docs in this project.

Claude Code should call exec (e.g. ls) and respond with your knowledge base. If it doesn't see the tool, restart Claude Code so it picks up the new MCP entry.

Live editing

When Claude Code writes to a doc, the desktop editor focuses on that doc and the change streams in live.

Open with AI

From the editor, Open with AI ▸ Claude dispatches the current file, folder, or project to Claude Code in one click. Three entry points: the editor header's sparkle icon, the sidebar's right-click menu (any file, folder, or empty space), and the macOS File → Open with AI menu.

Claude Code receives a short prompt asking it to open the target in Open Knowledge's web preview. The agent then grounds the rest of the conversation through the MCP tools, with no file attachment and no content over the wire. See Editor → Open with AI for the full UX.

Install detection hides the Claude row from the submenu when the desktop app isn't on the machine the OK server is running on. On remote-hosted setups (where the browser and server are on different machines) every agent renders and your browser's OS protocol-dispatch dialog decides what fires.