---
title: "Installation"
description: "Download and first-run Bi. Nothing to install — just download and run."
---

Bi runs straight after download — there is nothing else to install.

## Download

Head to the [download page](/en/download) and pick the build that matches your system:

- **Windows** (amd64 / arm64): double-click `bi-windows-*.exe`.
- **macOS / Linux** (amd64 / arm64): add execute permission, then run it.

## First run

Bi starts a local server and opens the workbench in your browser. On first run it creates a `.bi/` directory in the working directory:

- `config.json` — your local configuration
- `bi.db` — the SQLite database holding sessions and messages

The address shown in the UI is a local port. Bi has no server of its own: sessions, files and the database stay under the local `.bi/` directory. If you configure a remote model provider, conversation content is sent to the model endpoint you chose.

## Pick a model

Bi ships without a model. Before first use, configure at least one model provider (an OpenAI-compatible endpoint or a local model) — see [Model providers](/en/docs/configuration/providers).

## Verify the download

The download page lists a SHA-256 for every artifact. Use it to confirm the file is intact:

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

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

Compare the output with the value shown on the download page.