To use Sarah in your browser you only need an account and a Gemini API key. Nothing has to be installed. Chrome or Edge on a laptop works best for the microphone.

The desktop app needs nothing else installed. Install these only for the extras:

• Node.js 22 or newer (for the Developer agent / Claude Code) — nodejs.org/en/download. Check with node -v.
• Python 3.10 or newer (for the optional helper scripts) — python.org/downloads. Check with python3 --version (on Windows: py --version).

1. Open Google AI Studio → API keys and sign in with Google.
2. Press Create API key and copy it (it starts with AIza).
3. Paste it on the Connections page and press Save. The page tests it for you.

1. Create an account at fish.audio and add credit if you want the paid voices.
2. Open API keys, create a key (it starts with sk-fish-) and save it on Connections.
3. Choose a female or male voice on the Account page. English and Urdu each get a native voice.

1. In Google Cloud, create a project.
2. Enable the APIs you want: Gmail, Google Calendar, Google Drive, Google Search Console, YouTube Data, Google Analytics Data, Google Sheets.
3. Set up the OAuth consent screen (External) and add yourself as a test user.
4. Create an OAuth client ID of type Web application. Add the redirect URI shown on the Connections page.
5. Paste the client ID and secret on Connections, press Save, then Sign in with Google.
6. For Search Console, enter your property exactly as it appears in Search Console, e.g. https://example.com/ or sc-domain:example.com.

• Facebook & Instagram: in Graph API Explorer, get a long-lived Page access token, your Page ID and (for Instagram) your business account ID.
• X: in the developer portal, copy the bearer token; add a user access token to post.
• TikTok: create an app at TikTok for Developers and copy the client key, secret and a refresh token.
• Stripe: create a restricted, read-only key for the Finance agent.
• PayPal: in the PayPal developer dashboard, create a REST app, switch on Transaction search, and copy its Live client ID and secret.
• Bank account (Wise): in Wise → Settings → API tokens, create a token set to Read only. UK and EU accounts may ask you to approve statement access in the Wise app once.

The Developer agent reads and works on code projects on your computer using Claude Code, signed in with your own Claude Pro or Max plan.

1. Install Node.js (step 1), then in a terminal run npm install -g @anthropic-ai/claude-code.
2. Run claude once and log in with your Claude account.
3. The Developer agent then works through the Sarah desktop app on that computer. Anything that changes files waits for your approval.

A ChatGPT Plus subscription is for chatting on chatgpt.com and doesn't include API access, so other apps — including Sarah — can't use it. Sarah runs on your Gemini key. Support for OpenAI API keys (billed separately at platform.openai.com) is on the roadmap.

The Sarah desktop app links a computer to your account so your assistant can open apps and websites, switch Chrome profiles, list what's open, set the volume and play or pause music. Quitting apps, restarting and scripts always wait for your approval. Nothing else needs to be installed.

1. Download the app for your computer from the Download page.
2. Install it and open it. The first time, Mac may ask you to right-click → Open, and Windows may show “Windows protected your PC” → More info → Run anyway.
3. Sign in with your Sarah email and password. The computer shows as online on your Account page.
4. It keeps running in the menu bar (Mac) or system tray (Windows, Linux) and starts when you log in.

• Mac: the first time it controls an app, allow it under System Settings → Privacy & Security → Automation.
• Windows: volume and play/pause use the media keys, like your keyboard does.
• Linux: uses pactl for volume, playerctl for music and wmctrl to list windows (e.g. sudo apt install pulseaudio-utils wmctrl playerctl).

To unlink, choose Sign out in the app, or press Unlink on the Account page.