GA-Bagua Semantic KG — MCP Server

LLM semantic memory layer using Geometric Algebra and the I-Ching Bagua taxonomy.

Quick Start — One Command Setup

Option 1: One-liner (no npm needed)

Unix / macOS / Linux:

curl -fsSL https://raw.githubusercontent.com/trac41799/ga-bagua-semantic-kg/main/npm/setup.js | node -

Windows PowerShell:

Invoke-WebRequest -Uri "https://raw.githubusercontent.com/trac41799/ga-bagua-semantic-kg/main/npm/setup.js" -OutFile "$env:TEMP\ga-setup.js"; node "$env:TEMP\ga-setup.js"

The setup script will:

1. Download the correct binary for your platform
2. Auto-detect installed coding agent harnesses (Claude, OpenCode, Cursor, etc.)
3. Configure MCP server settings for each detected agent
4. Install the Bagua encoder skill into agent skill directories

Option 2: Via npm

npm install -g ga-semantics-mcp

Then run the setup wizard:

npx ga-semantics-setup

Or just start the MCP server:

npx ga-semantics-mcp

Option 3: Via Cargo (Rust toolchain)

cargo install ga-semantics-mcp

Then configure your LLM client manually (see Example Config below).

Supported Agents

The setup auto-detects and configures:

| Agent | Auto-Config | Auto-Skill | |-------|------------|------------| | Claude Desktop | MCP config | — | | Claude Code (CLI) | MCP config | CLAUDE.md + skill dir | | OpenCode | MCP config | .opencode/skills/ | | Cursor | MCP config | .cursor/skills/ | | Continue.dev | MCP config | .continue/skills/ | | Windsurf | MCP config | .windsurf/skills/ | | Aider | Manual instructions | — | | Cline (VS Code) | Manual instructions | .cline/skills/ | | Codex (OpenAI) | CLI fallback | — |

Example MCP Config

If configuring manually, add to your client's MCP config:

{
  "mcpServers": {
    "ga-bagua-semantic-kg": {
      "command": "ga-semantics-mcp"
    }
  }
}

See Delivery Guide for per-client config details.

Setup Options

--help              Show help
--quiet             Minimal output
--yes               Auto-confirm all prompts
--path <dir>        Binary install directory (default: ~/.ga-semantics/bin)
--skip-binary       Skip binary download (use existing)
--skip-config       Skip MCP config installation
--skip-skills       Skip skill file installation
--agents <list>     Only configure specific agents (comma-separated)
--list-agents       List detected agents without installing

What it does

• Encode any concept into an 8-number vector using the LLM's semantic understanding
• Classify relationships between concepts using WuXing cycle dynamics (generating, controlling, clarifying, etc.)
• Compute similarity with role-weighted metrics
• Solve analogies (A:B :: C:?) deterministically
• Persist knowledge graphs as JSON files with query and export capabilities
• 29 MCP tools for LLM-driven semantic reasoning

Documentation

• System Guide — complete technical reference
• Delivery Guide — configuration for every LLM client
• Encoding Skill — the LLM encoding protocol

License

MIT OR Apache-2.0