codenexus-ai 1.0.3

The context engine for AI coding agents - reduce tokens by 50-70%

CodeNexus AI

The context engine for AI coding agents

English | 한국어

---

What is CodeNexus?

CodeNexus is a local-first context engine that helps AI coding agents understand your codebase better. It builds a live dependency graph of your code and serves only the relevant context to AI agents, reducing token usage by 50-70% while improving code quality.

Key Features

• Local-first: Your code never leaves your machine
• Token reduction: Save 50-70% on AI API costs
• Multi-language: Python, JavaScript, TypeScript support
• MCP compatible: Works with Claude Code, Cursor, and other AI agents
• Fast indexing: SQLite-based dependency graph

---

Quick Start

Installation

pip install codenexus-ai

Basic Usage

# Index your project
codenexus index

# Search for context
codenexus search "authentication middleware"

# Run context pipeline
codenexus pipeline "fix login bug"

# Check index status
codenexus status

---

How It Works

Your Codebase
     ↓
[CodeNexus Indexer]
     ↓
┌─────────────────────────────────────┐
│  Dependency Graph (SQLite + FTS5)   │
│  - Functions, Classes, Imports      │
│  - Call relationships               │
│  - Type information                 │
└─────────────────────────────────────┘
     ↓
[Context Capsule Generator]
     ↓
┌─────────────────────────────────────┐
│  Optimized Context for AI Agent     │
│  - Pivot files: Full source         │
│  - Supporting: Skeleton only        │
│  - Token budget: Respected          │
└─────────────────────────────────────┘
     ↓
AI Agent (Claude Code, Cursor, etc.)

---

Claude Code Integration

CodeNexus works with Claude Code to reduce token usage for AI coding agents.

Setup

1. Install CodeNexus

pip install codenexus-ai

2. Edit .claude.json

File location:

• macOS/Linux: ~/.claude.json
• Windows: C:\Users\your-username\.claude.json

3. Add configuration

Add the following to your ~/.claude.json:

{
  "mcpServers": {
    "codenexus": {
      "command": "codenexus",
      "args": ["serve", "-w", "/path/to/your/project"]
    }
  }
}

⚠️ Important: /path/to/your/project should be the path to the project where you want to use CodeNexus.

Path Examples

❌ Wrong:

"args": ["serve", "-w", "C:\\Users\\username\\.codenexus"]

→ CodeNexus config directory (incorrect)

✅ Correct:

"args": ["serve", "-w", "C:\\Users\\username\\projects\\my-app"]

→ Project directory where you want to use CodeNexus

OS-specific Path Examples

Windows:

{
  "mcpServers": {
    "codenexus": {
      "command": "codenexus",
      "args": ["serve", "-w", "C:\\Users\\username\\projects\\my-app"]
    }
  }
}

macOS/Linux:

{
  "mcpServers": {
    "codenexus": {
      "command": "codenexus",
      "args": ["serve", "-w", "/home/username/projects/my-app"]
    }
  }
}

Multiple Projects

To use CodeNexus with multiple projects, add configuration for each project:

{
  "projects": {
    "C:\\Users\\username\\projects\\app1": {
      "mcpServers": {
        "codenexus": {
          "command": "codenexus",
          "args": ["serve", "-w", "C:\\Users\\username\\projects\\app1"]
        }
      }
    },
    "C:\\Users\\username\\projects\\app2": {
      "mcpServers": {
        "codenexus": {
          "command": "codenexus",
          "args": ["serve", "-w", "C:\\Users\\username\\projects\\app2"]
        }
      }
    }
  }
}

Verify Setup

1. Save .claude.json
2. Restart Claude Code
3. Run Claude in your project directory

cd C:\Users\username\projects\my-app
claude

Troubleshooting

If MCP server doesn't connect:

1. Check if CodeNexus is installed:

   pip show codenexus-ai
   

2. Test code execution:

   codenexus --version
   

3. Verify the path is correct (watch for quotes)

---

Supported Languages

Language	Status
Python	✅ Full support
JavaScript	✅ Full support
TypeScript	✅ Full support
Go	✅ Full support
Rust	✅ Full support
Java	✅ Full support
C#	✅ Full support

---

Other AI Agent Integration

CodeNexus works with various AI coding agents. Here's how to integrate with other popular tools.

OpenClaw Integration

OpenClaw is a personal AI assistant that connects to WhatsApp, Telegram, Slack, Discord, and more.

Setup

OpenClaw doesn't support MCP directly. Use CodeNexus CLI commands through OpenClaw's skill system.

1. Install CodeNexus:

pip install codenexus-ai

2. Create OpenClaw skill:

Create ~/.openclaw/workspace/skills/codenexus/SKILL.md:

---
name: codenexus
description: Search and analyze code using CodeNexus
allowed_tools:
  - bash
---

# CodeNexus Skill

Use CodeNexus to search and analyze code in the workspace.

## Commands

- `codenexus index` - Index the workspace
- `codenexus search "query"` - Search for code
- `codenexus pipeline "task"` - Get context for a task

3. Usage in OpenClaw:

/codenexus search "authentication middleware"
/codenexus pipeline "fix login bug"

Hermes Agent Integration

Hermes Agent is a self-improving AI agent by Nous Research.

Setup

Hermes supports MCP servers. Configure CodeNexus as an MCP server.

1. Install CodeNexus:

pip install codenexus-ai

2. Add MCP server to Hermes:

hermes mcp add codenexus -- codenexus serve -w /path/to/your/project

Or add to ~/.hermes/config.yaml:

mcp_servers:
  codenexus:
    command: codenexus
    args:
      - serve
      - -w
      - /path/to/your/project

3. Usage in Hermes:

/hermes search "authentication middleware"
/hermes pipeline "fix login bug"

Other Agents

Any agent that supports CLI commands can use CodeNexus:

# Direct CLI usage
codenexus index
codenexus search "query"
codenexus pipeline "task"
codenexus status

# JSON output for programmatic use
codenexus search "query" --json

---

Token Savings Example

Before CodeNexus:

8,247 tokens per query

After CodeNexus:

2,140 tokens per query (74% reduction)

---

Roadmap

• Tree-sitter integration for better parsing
• Graph centrality (PageRank) for better ranking
• Local LLM support for additional savings
• Multi-repo workspace support
• VS Code extension

---

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the repository
2. Create your feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

---

License

This project is licensed under the MIT License - see the LICENSE file for details.

---

Acknowledgments

• Inspired by vexp - The original context engine
• Built with Python, SQLite, and MCP
• Thanks to all contributors

---

Support

If you find CodeNexus useful, consider supporting the project:

---

Connect your code. Empower your AI.