Skip to main content

Kenaz documentation

Welcome to Kenaz, by Kameas AI: a local-first platform for AI-assisted work. Your files, keys and history stay on your machine. What leaves it is the model calls you make to the provider you choose, plus the network requests listed under What connects to the network. Kameas's cloud service, Fleet, is optional.

Kenaz is three products:

  • Kenaz, the desktop app. It creates and runs workbenches: virtual machines with the Harness and other tools inside. On macOS (Apple Silicon) a workbench is a Tart VM. On Linux, Kenaz runs code-server as a process with no VM. There is no Windows build. Free to use, no account needed.
  • Kenaz Harness, open source (Apache-2.0). The AI agent app. Bring your own provider keys (Anthropic, OpenAI, Azure OpenAI, Google Gemini, OpenRouter, Bedrock, Ollama, or any OpenAI-compatible endpoint), run tools and MCP servers, manage sessions, context and memory, and keep a hash-chained local audit log. It runs inside a workbench, or on its own on macOS, Windows and Linux. Free to use, no account needed.
  • Kenaz Fleet, hosted by Kameas AI, paid. The cloud control plane for individuals and teams: sync, org and team management, workbench and Harness policy, shared context, audit archive, emergency lockdown and SSO.

What's free and what's paid​

The desktop app, its workbenches and the Harness are fully usable on your machine with no account and no plan. Everything that needs Fleet is paid. Signing in without a plan gives you no Fleet features, and nothing local stops working. Plans & Billing lists what each plan adds.

Source code​

The Harness is open source: github.com/kameas-ai/kenaz-harness, under Apache-2.0. The Kenaz desktop app and the workbench images are not public today. Fleet is a closed-source service that Kameas AI operates.

What connects to the network​

WhatWhenWhere it goesHow to turn it off
Model callsEach turn you sendThe AI provider you configured, directlyUse a local provider such as Ollama
Harness update checkAt launch and every few hoursdownloads.kameas.ai (the release manifest; nothing about you is sent)Settings → Updates, turn off the automatic check
Desktop app update checkAt most once a day, in releases that have itdownloads.kameas.ai (the release manifest)The Check for updates automatically setting
Tart install (macOS)When you click Install on the Tart sheetGitHub releasesInstall Tart yourself
Workbench images (macOS)When a workbench needs an image that isn't in the local cachedownloads.kameas.ai—
FleetOnly after you sign inKameas Fleet: sign-in, enrollment, policy and config polling, and the sync features your plan and settings turn onSign out
Telemetry and crash reportsOnly after you opt inKameas FleetOff by default
Connectors and MCP serversWhen you add one and the model uses itThat serviceRemove the connector

Where to start​

Getting help​