Skip to main content

First run

This page covers the first launch of Kenaz and what has to be in place before a workbench can open. Install Kenaz first.

1. Sign in, or skip​

The first screen says:

Sign in to connect your Kameas account, or continue in local mode and sign in later from Settings.

  • Sign in opens your web browser to finish signing in to your Kameas account. The app shows "Complete sign-in in your browser…" while it waits.
  • Skip for now continues without an account. Kenaz doesn't ask again; you can sign in later from Settings → Account.

Kenaz keeps your sign-in tokens in your operating system's keychain.

2. Pick your role​

Next, Kenaz asks about your role (for example engineer, data scientist, product, research). Your answer decides which screen Kenaz opens on and how the sidebar is arranged. Change it any time in Settings → Role & persona.

3. The main window​

The main window has a sidebar. The entries you need for workbenches are:

  • Workbenches: your workbenches, their state, and the buttons to manage them.
  • Profiles: the recipes workbenches are created from.
  • App Library: the apps a profile can include, and MCP connectors.
  • Egress: the network destinations a workbench may reach.

When you have no profiles, Kenaz creates one for you, Daily Driver. See Create and run workbenches.

Tart​

On macOS, workbenches run in Tart virtual machines. The first time you open a workbench, Kenaz checks for Tart in ~/.kenaz/bin, /opt/homebrew/bin, /usr/local/bin, /usr/bin and your shell's PATH.

If Tart is missing, or older than the version Kenaz is tested against, a sheet titled Tart is needed to run workbenches appears:

  • Install downloads the pinned Tart release from Tart's GitHub releases, checks its SHA-256 checksum, and installs it under ~/.kenaz (~/.kenaz/tart.app, with ~/.kenaz/bin/tart pointing to it). No administrator password is needed. A copy of Tart's license is saved as ~/.kenaz/tart.LICENSE. If you already have an older Tart, Kenaz leaves it alone and installs its own copy next to it.
  • Not now closes the sheet. The rest of Kenaz keeps working.
  • To install Tart yourself, run brew install tart, then click I already installed it — re-check.

Tart's licence​

Tart is a separate program with its own licence; Kenaz doesn't include it, and installs it only when you click Install. Tart was made by Cirrus Labs, which has joined OpenAI; its home is now github.com/openai/tart.

  • The release Kenaz installs today, 2.32.1, is under Cirrus Labs' Fair Source licence. Read the copy saved as ~/.kenaz/tart.LICENSE for its terms, including any limits on commercial use.
  • Kenaz is moving to a current Tart release. Current releases are licensed under FSL-1.1-ALv2 (the Functional Source License): use is permitted except for a competing product, and each release becomes Apache-2.0 two years after it is published. This page will say when the Kenaz release that installs it is out.

If you install Tart yourself, the licence of the version you install applies.

Workbench images​

On macOS, each workbench is cloned from a workbench image. Kenaz looks for an image in this order:

  1. Its local image cache, ~/.kenaz/workbench/images.
  2. The downloads site, downloads.kameas.ai. Kenaz checks the image's signature and checksum before using it.
  3. A local build with Nix. This is for developers only; it needs developer mode (KENAZ_DEV_MODE=1), a Kenaz source checkout and a registered Nix builder.

Only profiles whose apps and egress tier match a published image can use steps 1 and 2. A custom combination always needs step 3. The profile editor warns you when a profile doesn't match a published image.

:::caution No images are published yet At the moment no workbench images are published to the downloads site. On a release build, opening a workbench on macOS stops with an error that begins:

imagecache: this workbench profile has no published image

Until images are published, opening a workbench needs the developer setup in step 3. :::

Linux hosts​

On Linux, Kenaz doesn't create a virtual machine or use workbench images. Each workbench is a code-server process running directly on your computer, listening on 127.0.0.1 with authentication turned off. The profile's apps, resources and network settings aren't applied, and there's no isolation between the workbench and the rest of your computer.

Install code-server and make sure it's on your PATH before you open a workbench, for example with npm i -g code-server. If it's missing, opening a workbench fails with:

hostvm: code-server not found on PATH

Next​

Create and run workbenches.