Haris
A security-first desktop AI assistant for Windows that talks, listens and acts on your PC without trusting the model, in English or Arabic (right to left). Open source under the MIT license. Now at version 0.6.0, which adds a Stop button that ends a reply mid-answer.
Role: Creator and maintainer
- Every tool call passes a permission gate: read-only and low-risk tools run and are logged; anything that opens apps, changes files, touches the clipboard, screen or camera, or powers off waits for your click
- Approvals come only from your click: never from voice, never from another AI model
- Choose the engine: Claude, Ollama or LM Studio, any OpenAI-compatible API, or Ollama on a free Google Colab GPU
- The core binds to 127.0.0.1 only, keys stay in Windows Credential Manager, memory is encrypted with AES-256-GCM, and it installs per user with no administrator rights
- Java 21
- Spring Boot 4
- Vite
- Docker
- Vosk
- Windows Credential Manager
- Source code on GitHub (opens in a new tab)
- Visit the website (opens in a new tab)
- Download version 0.6.0 (opens in a new tab)
The installer is not code-signed yet: check the SHA-256 in the release notes and unblock the zip.


















