Skip to content

Quick Start

This guide takes you from zero to a working Omnodex installation with live risk detection.

Run every step on the host where your agent runs. On Windows, that is PowerShell for desktop apps and agents you start from PowerShell, or a WSL terminal for agents you run inside WSL. See Windows and WSL.

  1. Install Omnodex from source:

    Terminal window
    git clone https://github.com/omnodex/omnodex.git
    cd omnodex
    npm install
    npx tsc -b
    npm install -g ./packages/cli

    On Windows PowerShell, use .\packages\cli in the last command. Check the install with omnodex --version.

  2. Point hooks at your build.

    Create omnodex-config.json in your Omnodex home (~/.omnodex/, or C:\Users\<you>\.omnodex\ on Windows) with the paths of the hook handlers in your clone. The Installation page has the exact file. Skip this step if you only use Cowork.

  3. Set up your agent:

    From your project directory:

    Terminal window
    omnodex install claude-code

    Then start a new Claude Code session in the project. See the Claude Code guide.

    omnodex install ends by offering to connect to the hosted dashboard. Continue with the next step, or press Ctrl+C to skip it.

  4. Connect to the hosted dashboard (optional):

    Terminal window
    omnodex connect

    The CLI shows a short code and a URL. Open the URL, sign in or create an account at dashboard.omnodex.com, and enter the code. The connection completes automatically, and your API token and sync passphrase are saved for future sessions. The passphrase is transferred end-to-end encrypted, so the server never sees it.

    Skip this step if you only want the local dashboard.

  5. Use your agent normally. Omnodex records activity in the background without changing what the agent does.

  6. View results:

    Terminal window
    omnodex dashboard

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

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

  • A timeline of the tool calls and actions the agent took
  • Risk indicators for concerning behavior, such as credential exposure, sensitive file access, or unexpected network calls
  • Details for each event: what was called, with which parameters, and how it ended