Skip to content

DeepSeek Harness for your desktop. ​

Work in your local projects with your choice of models and a plugin-based agent.

Harness Desktop conversation workspace before a model is configured

Choose your model. ​

Configure DeepSeek, other supported providers, or an OpenAI-compatible gateway. Your provider handles inference and billing.

Model setup and authentication

Harness Desktop Models settings with provider configuration controls

A desktop app with room to extend. ​

Harness owns the agent runtime. Tauri supplies the native window and operating-system integration.

Native where it matters ​

System WebView, native menus, keyboard shortcuts, runtime supervision, and diagnostic export. No bundled Electron or Chromium.

Desktop architecture

Plugins at the core ​

Use Cordis plugins, Agent presets, skills, and MCP integrations. Optional Codex and Claude Code bundles add delegated agents after setup.

Plugin development

Start with a local project. ​

  1. Install. Choose the macOS Apple Silicon or Windows x64 installer from the release assets.
  2. Connect. Configure your model provider in Settings, using its supported authentication method.
  3. Work. Choose a workspace and an Agent preset, then describe your task.

Before you install ​

This is a community preview, not an official DeepSeek product. macOS previews are ad-hoc signed and not notarized; Windows previews are unsigned. Check each release's platform notes and bundled version.

Local execution is not offline inference. Plugins and external agents run code; review permissions before granting workspace access. Automatic updates are not configured.

Download