Installation
Tarsk runs as a local server that serves both the REST API and the React frontend from a single process. There are two ways to run it.
Option 1: npx (recommended)
Section titled “Option 1: npx (recommended)”No installation required. Run the following command in any terminal:
npx tarskTarsk starts on port 641 and opens http://localhost:641 in your browser automatically. The server stays running in that terminal session — press Ctrl+C to stop it.
To always use the latest version:
npx tarsk@latestOption 2: Electron Desktop App
Section titled “Option 2: Electron Desktop App”Download the desktop app from github.com/damiant/tarsk/releases. Prebuilt binaries are available for:
- macOS — DMG (Apple Silicon & Intel) or ZIP
- Windows — NSIS installer or portable EXE
- Linux — AppImage or .deb
The Electron app bundles the same CLI server and starts it automatically. No separate terminal is needed.
System Requirements
Section titled “System Requirements”| Requirement | Notes |
|---|---|
| git | Must be installed and on your PATH. Tarsk uses git to clone repositories and manage branches. Download from git-scm.com. |
| Node.js 18+ or Bun | Required for npx tarsk. The Electron app bundles its own runtime. |
| gh CLI (optional) | Required only for creating GitHub repositories and pull requests from within Tarsk. Install from cli.github.com. |
Tarsk verifies that git is installed during the onboarding wizard on first launch and will show a warning with a download link if it is missing.
Where Data Is Stored
Section titled “Where Data Is Stored”All Tarsk data lives in your user data directory:
| Platform | Path |
|---|---|
| macOS | ~/Library/Application Support/Tarsk/ |
| Windows | %APPDATA%\Tarsk\ |
| Linux | ~/.config/Tarsk/ |
Inside this directory:
tarsk.db— SQLite database (projects, threads, conversation history, settings)data/{threadId}/— one directory per thread, containing the full git clone for that thread
This means each thread is a real, independent filesystem copy of your repository. You can open any thread directory directly in your IDE or file manager.
Next Steps
Section titled “Next Steps”Continue to the Quickstart to connect your first repository and start working with the AI agent.