Xknow-CLI 🧠

AI-First Knowledge Management for OpenClaw Users - Based on Karpathy LLM Knowledge Bases concept

Xknow-CLI is an automated knowledge management tool designed for AI agents and their owners. It follows the principle: "LLM writes and maintains the Wiki, Humans ask and discover."

Key Features

• 🚀 Incremental Compilation: Only processes new or modified raw data using MD5 hashing.
• 🔍 Context-Aware Q&A: Leverages the full KB context for deep queries (beyond standard RAG limitations).
• 🛡️ AI Health Check: Automatically audits your Wiki for orphans, duplicates, and missing connections.
• 🔌 MCP Support: Native Model Context Protocol server support for integration with AI agents (Claude Desktop, OpenClaw, etc.).
• 🖇️ OpenClaw Integration: Automatically bridges your existing OpenClaw credentials and model settings (found in ~/.openclaw/openclaw.json).
• 📂 Obsidian Friendly: Uses Obsidian as the IDE/frontend for your knowledge. Your data stays local.

Model Context Protocol (MCP) Integration

Xknow-CLI acts as an MCP server, allowing AI agents to directly use your knowledge base as a set of tools.

Setup in Claude Desktop

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "xknow": {
      "command": "node",
      "args": ["/PATH/TO/Xknow/bin/xknow-mcp.js"]
    }
  }
}

Available MCP Tools

• xknow_query: Intelligent Q&A on your Wiki.
• xknow_search: Fast local keyword lookup.
• xknow_compile: Transform raw chaos into ordered knowledge.
• xknow_ingest: Add new sources to your knowledge base.
• xknow_lint: Perform an AI librarian health audit.

Configuration & Credentials

Xknow-CLI is designed to work with zero configuration for OpenClaw users. It automatically looks for your API keys and models in: ~/.openclaw/openclaw.json

Manual Setup (Without OpenClaw)

If you're not an OpenClaw user, you can set the following environment variables:

export OPENAI_API_KEY="your-api-key"
export OPENAI_BASE_URL="https://api.openai.com/v1"
export OPENAI_MODEL="gpt-4o"

You can verify your configuration at any time by running:

xknow-cli config --list

~/Obsidian/Xknow-Wiki/       # Your Knowledge Base (Local)
├── raw/                     # Chaos: Your raw input (articles, papers, notes)
├── wiki/                    # Order: LLM-compiled structured wiki pages
├── INDEX.md                 # Global navigation index
└── .xknow-history.json      # Compilation state & history

Quick Start

1. Install

Via NPM:

npm install -g xknow-cli

Via PNPM:

pnpm add -g xknow-cli

For Development:

pnpm install && pnpm link --global

2. Initialize

xknow-cli init

3. Feed the Chaos

Drop your raw Markdown, PDF (via extract), or text files into ~/Obsidian/Xknow-Wiki/raw/notes/ (or other subdirs).

4. Compile to Order

xknow-cli compile

5. Ask Anything (AI Synthesis)

xknow-cli query "What are the core design patterns used in my recent projects?" --format slides --save

6. Fast Search (Local/Offline)

xknow-cli search "LLM"

7. Diagnose Setup

xknow-cli doctor

Karpathy Methodology

According to Andrej Karpathy's Twitter insights:

1. LLM as Editor: Writing and maintaining the Wiki is the LLM's job, not yours.
2. Beyond RAG: For KBs within 100k tokens, providing the full context to a large LLM outperforms traditional RAG.
3. Autonomous Links: The LLM automatically generates [[Wikilinks]] and back-references between concepts.

Development

• Built with Node.js 18+ (ES Modules)
• Uses commander, chalk, ora, openai, globby
• CI/CD: Automated NPM release via GitHub Actions on v* tag.

License

MIT