Skip to main content

Getting started

This page takes you from a fresh machine to your first open workspace.

:::note Preview A signed, downloadable Workbench build isn't out yet. For now you run it by building from source — the steps below cover the prerequisites and the shape of that. When the download lands it'll be on the download page and this page will switch to "download and open." :::

1. Check your machine​

The Workbench runs an isolated Linux environment, so it needs a virtualization substrate your OS already provides.

macOS​

  • macOS 13 (Ventura) or newer, on Apple Silicon (M-series). Intel Macs aren't supported yet.
  • Tart — the lightweight VM tool the Workbench uses: 
    brew install cirruslabs/cli/tart
  • Xcode Command Line Tools: 
    xcode-select --install

Windows​

  • Windows 11 (Windows 10 can work with WSLg installed manually).
  • WSL2 enabled: 
    wsl --install
    Restart if prompted, then confirm WSL2 is the default: 
    wsl --set-default-version 2

You do not need Docker to run the Workbench. (Docker is only used by maintainers when building the guest image.)

2. Get the Workbench​

Until the download is published, build it from source. You'll need Go 1.24+ in addition to the prerequisites above.

  1. Clone the repository: 
    git clone https://github.com/kameas-ai/kenaz-workbench
    cd kenaz-workbench
  2. Follow the build steps in the repository's README.md. They produce the launcher — the app you'll run — and prepare the guest image for your platform.

See Build from source for the detailed, platform-specific build reference.

3. Open your first workspace​

  1. Run the launcher. Started with no arguments, it opens a small control page in your browser at http://127.0.0.1:8181. (If your browser doesn't open automatically, visit that address.)
  2. Open the workspace. On the control page, start the workbench. The launcher provisions your sandbox — a Tart VM on macOS, a WSL2 distribution on Windows — and brings it up. The first launch is the slowest (it prepares the image); later launches are quick.
  3. Start working. Once the workspace is up, the Kenaz Harness is your main surface — pick a provider, and you're ready to chat and run tools, all inside the sandbox. See Using the Workbench.
  4. Close it when you're done. Closing the workspace tears the sandbox down cleanly.

:::tip First time in the Harness? The Harness needs at least one AI provider before it can do anything. The Harness quickstart walks through adding a provider key — that flow is the same inside the Workbench. :::

Trouble?​

If the launcher won't start, the workspace won't open, or the control page is blank, see Troubleshooting & FAQ.