@aaronsb/kg-cli

CLI and MCP server for interacting with Knowledge Graph System deployments.

Installation

Global install (requires sudo or npm configured for user-local):

npm install -g @aaronsb/kg-cli

User-local install (no sudo required):

npm install -g @aaronsb/kg-cli --prefix ~/.local

Then ensure ~/.local/bin is in your PATH.

Run without installing:

npx @aaronsb/kg-cli health
npx @aaronsb/kg-cli search "query"

This installs two commands:

• kg - Command-line interface for the knowledge graph
• kg-mcp-server - MCP server for AI assistant integration

Quick Start

# Configure your knowledge graph endpoint
kg config set api.url https://kg.example.com/api

# Login
kg login

# Check health
kg health

# Ingest a document
kg ingest document.pdf

# Search concepts
kg search "machine learning"

CLI Commands

| Command | Description | |---------|-------------| | kg health | Check API connection | | kg login | Authenticate with the API | | kg logout | Clear saved credentials | | kg config | View/edit configuration | | kg ingest <file> | Submit document for extraction | | kg search <query> | Search concepts | | kg jobs | List extraction jobs | | kg artifact | Manage artifacts | | kg document | Manage documents |

Run kg --help or kg <command> --help for detailed usage.

MCP Server

The MCP server allows AI assistants (like Claude) to interact with your knowledge graph.

Claude Desktop Configuration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or ~/.config/Claude/claude_desktop_config.json (Linux):

{
  "mcpServers": {
    "knowledge-graph": {
      "command": "npx",
      "args": ["-p", "@aaronsb/kg-cli", "kg-mcp-server"],
      "env": {
        "KG_API_URL": "https://kg.example.com/api",
        "KG_API_TOKEN": "your-token-here"
      }
    }
  }
}

Or if installed globally/user-local (simpler):

{
  "mcpServers": {
    "knowledge-graph": {
      "command": "kg-mcp-server",
      "env": {
        "KG_API_URL": "https://kg.example.com/api",
        "KG_API_TOKEN": "your-token-here"
      }
    }
  }
}

Available MCP Tools

• search_concepts - Semantic search across the knowledge graph
• get_concept - Get details about a specific concept
• list_sources - List ingested documents
• get_relationships - Find connections between concepts
• ingest_document - Submit new documents (with approval)

Configuration

Configuration is stored in ~/.config/kg/config.json:

{
  "api": {
    "url": "https://kg.example.com/api"
  },
  "auth": {
    "token": "..."
  }
}

Or use environment variables:

• KG_API_URL - API endpoint
• KG_API_TOKEN - Authentication token

Requirements

• Node.js 18+
• A running Knowledge Graph System instance

Shell Alias (optional)

If using npx, add an alias to your shell profile:

alias kg='npx @aaronsb/kg-cli'

Links

• Knowledge Graph System
• Documentation
• Issues

License

MIT