---
title: "Gemini CLI setup"
description: "Configure the Gemini-compatible endpoint, key and model for terminal use."
lastUpdated: 2026-10-22
---

Install Gemini CLI following its [official instructions](https://geminicli.com/docs/get-started/installation/). Use a supported Node.js version for the release you install.

## Configure the endpoint

Set these variables in the terminal that will launch Gemini CLI:

```bash
export GEMINI_API_KEY="YOUR_API_KEY"
export GOOGLE_GEMINI_BASE_URL="https://infistar.ai"
export GEMINI_API_KEY_AUTH_MECHANISM="bearer"
export GEMINI_MODEL="GEMINI_MODEL_ID"
gemini
```

In PowerShell, use `$env:NAME = "value"` for each variable instead. Replace both placeholders. The base address has no `/v1` or `/v1beta` suffix; the client supplies the protocol path.

Select API-key authentication when prompted. Newer releases may also expose a gateway authentication flow; use the [configuration for your installed release](https://geminicli.com/docs/get-started/configuration/). Upstream supports a custom Gemini base address and Bearer authentication, but Infistar has not tested every CLI release and model combination.

## Verify and save

Start with: “Reply with configuration successful. Do not read or modify files or run commands.” Check the result before allowing project actions. After verification, the variables can be stored in your private `~/.gemini/.env`; do not commit the file.

For 401, check the key and Bearer setting. For 404, check the root address. For an unavailable model, select an exact Gemini-capable ID from your key's model list.