🧠 OMNIA: AST Codebase Memory MCP

Deterministic AST Symbol Graphs & 94% Token Reduction for Claude Desktop, Cursor & OpenHands

Stop stuffing entire 5,000-line files into your AI context window.
OMNIA parses your repository into an AST symbol graph, serving surgical caller/callee slices to AI coding agents in <16ms.

⚡ 10-Second Claude Setup • 📊 Token Compression Benchmarks • 🛠️ Available MCP Tools

---

🚀 10-Second Setup in Claude Desktop / Cursor

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "omnia-memory": {
      "command": "python",
      "args": ["-m", "core.omnia_server", "--workspace", "C:/path/to/repo"]
    }
  }
}

---

📊 Token Compression Benchmarks

Query Type	Traditional Full File Read	OMNIA AST Symbol Retrieval	Token Reduction
Single Function Call	2,840 tokens	148 tokens	-94.8%
Class Interface Audit	8,200 tokens	612 tokens	-92.5%
Dependency Flow Trace	14,500 tokens	890 tokens	-93.8%

---

🛠️ Core Capabilities

• Deterministic Symbol Map: Fast AST indexing across Python, TypeScript, C++, and Go.
• Caller/Callee Dependency Graphs: Instant resolution of upstream and downstream references.
• Relational Memory Storage: High-performance SQLite WAL backend with zero external database dependencies.