---
title: "Install and first run"
description: "Download, unzip, configure a model — three steps to start using Bi."
lang: en
---

Bi ships as an archive or an installer. There is no dependency to install and no separate runtime.

## Download

Go to the [download page](/download) and pick the build for your system. Each platform offers x64 and ARM64:

- **Windows**: unzip, then run `bi-windows-*.exe`. Unzip first — launching from inside the archive prevents `ocr/` from loading.
- **macOS**: the `.dmg` installer is recommended; mount it and drag Bi into Applications. A portable `.zip` is also available.
- **Linux**: unzip and run the executable for your architecture.

`ocr/` is the document recognition component and must sit next to the executable. Removing it only disables text recognition for scans; everything else keeps working.

## First run

Bi runs on your machine and opens the workbench in your browser. The platforms differ slightly:

- **Windows** launches in GUI mode and stays in the system tray.
- **macOS** provides a menu bar icon that opens the workbench, workspace, config directory and logs.
- **Linux** runs as a service; the workbench is reached in the browser.

Data lives in the `.bi` directory next to the application: settings and model keys, sessions, the local database and logs. Bi has no server of its own.

## Configure a model

Bi does not ship a model. Before first use, add at least one model provider in Settings — a cloud API, or a local Ollama / LM Studio instance. See [Add a model](/docs/configuration/providers).

## Verify the download (optional)

The download page lists a SHA-256 for every file. Run the command for your system:

```bash
# macOS / Linux
shasum -a 256 bi-linux-amd64.zip

# Windows PowerShell
Get-FileHash .\bi-windows-amd64.zip -Algorithm SHA256
```

Compare the output with the value on the download page.