Quick Start

This guide gets you from zero to a working Omnodex installation with live risk detection. The whole process takes under five minutes.

Prerequisites

• Node.js 24+ (required for the built-in SQLite module)
• npm (comes with Node.js)
• An AI coding agent: Claude Code, OpenAI Codex, or Google Antigravity

Install

1. Install Omnodex:

   npm (recommended)

   npm install -g omnodex

   From source

   git clone https://github.com/omnodex/omnodex.git
   cd omnodex
   npm install && npx tsc -b

   When running from source, use npx omnodex instead of omnodex in all commands below.

2. Install the hook into your agent:

   Claude Code

   omnodex install claude-code

   This registers Omnodex as a Claude Code hook. The next time Claude Code runs, all tool calls will be traced automatically.

   Codex

   omnodex install codex

   This creates a .codex/hooks.json configuration that routes Codex events through Omnodex.

   Antigravity

   omnodex install antigravity

   This creates .agents/hooks.json for hook-based tracing. Add --mcp to also install the MCP proxy. Works across CLI, Desktop App, and IDE extensions.

3. Run your agent normally. There is nothing else to configure. Omnodex intercepts in the background with zero impact on the agent’s behavior.

4. View results:

   omnodex dashboard

   Opens the Omnodex dashboard at http://localhost:7890 with a real-time view of all traced events and risk findings.

What You Should See

After running your agent with Omnodex installed, the dashboard shows:

• A timeline of all tool calls and actions the agent took
• Risk indicators flagging any concerning behavior (credential exposure, sensitive file access, unexpected network calls)
• Full context for each event - what was called, with what parameters, and what it returned

Next Steps

• Installation - Detailed installation options and configuration
• Claude Code Setup - Deep dive on the Claude Code integration
• Running Detection - Configure and customize risk detection rules