TokenOS

[](https://badge.fury.io/js/tokenos)

> Local-first codebase graph intelligence for AI assistants — powered by SQLite, ts-morph, and Ollama.

TokenOS is a Model Context Protocol (MCP) server that statically analyses your TypeScript/TSX codebase, stores it as a structural dependency graph in SQLite, optionally enriches nodes with semantic embeddings via Ollama, and exposes high-precision query tools for AI coding assistants like Claude, Cursor, or any MCP-compatible client.

The goal: When you start a new chat, the AI already knows your codebase structure. No more "let me analyze all files first" — it queries the graph and gets exactly what it needs, saving tokens and compute.

---

• Quick Start
• Commands
• Database Location
• MCP Client Configuration
• MCP Tools
• Node Types
• Edge Types
• Importance Scoring
• Semantic Meta Enrichment
• Conversation Memory
• Visualization Dashboard
• Changing the Embedding Model
• Architecture
• Configuration Reference
• Prerequisites
• Tech Stack
• Limitations
• License

---

Run TokenOS on any codebase instantly via npx. No installation, setup, or config files required!

# In your project folder
npx -y tokenos .

# Or for a specific path
npx -y tokenos /absolute/path/to/project

That's it. The server will:

1. Parse all .ts/.tsx files via ts-morph AST analysis
2. Extract nodes (functions, classes, components, interfaces, types, enums, routes, variables, imports)
3. Extract edges (CALLS, IMPORTS, EXPORTS, EX