Settings

Open them with in Elissa's window or from the tray icon. Changes save on their own. Sections that don't apply to your system don't show up.

General and AI

• Language: Spanish or English, or automatic.
• Assistant's name and your name.
• Artificial intelligence: shows the provider and model in use. Change provider or model opens the setup again.

Voice

• Voice and Speed (0.8× to 1.6×).
• Speech recognition: how she understands you (details).
• Wake her by saying…: "Hey Elissa", offline.
• Interrupt her by talking: best with headphones.
• Shortcut to talk: click it and press your combination.

All voice options

Option	Default	Notes
Voice	Salomé (es) / Ava (en)	Natural Edge voices, plus offline system voices
Speed	1.10×	0.8× to 1.6×
Speech recognition	Automatic	Groq, OpenAI, On your PC, Your own server
Local model	Fast · 60 MB	Or More accurate · 190 MB. Only for On your PC
Wake her by saying…	Off	"Hola / Oye Elissa" with a Spanish voice; "Hey / Hi / Okay Elissa" with an English one. Downloads 40 MB once
Interrupt her by talking	Off
Turn other apps down while we talk	20 %	10 %, 20 %, 35 % or not at all. On macOS, Spotify and Music
Continuous conversation	On	She listens again after each answer
Always on top	On
Start with the system	—
Shortcut to talk	Ctrl+Alt+Space	Must include Ctrl, Alt or the Windows/Command key

Sounds

• Elissa's sounds on or off, and their volume.
• Your own sounds: choose Open sounds folder, drop a .wav, .mp3, .ogg or .m4a named like the sound you want to replace (listen, done, greet, error, approval…) and choose Reload.

Email

1. Turn on Let her read my email (and Let her send email if you want).
2. Type your address and an app password, not your normal password.
3. Test and save email.

The server is detected for Gmail, Outlook, Yahoo and iCloud. For Gmail, create the password at myaccount.google.com/apppasswords.

She never deletes email or marks it as read, and every email she sends is shown to you first.

Integrations (MCP)

Pick one in Add:

	You need
GitHub	A personal access token
Jira and Confluence	Node.js; you sign in in your browser
Google Calendar	Node.js and your Google Cloud OAuth file
Other MCP server…	Its URL or command

Reading runs freely; changing data asks first.

Warning

Only add servers you trust: a local one is a program running on your computer.

Spotify

1. Create a free app at developer.spotify.com/dashboard.
2. Add the Redirect URI http://127.0.0.1/spotify/callback and tick Web API.
3. Paste its Client ID and Connect Spotify.

Playing what you ask for needs Spotify Premium. Pause, next and "what's playing?" work without connecting.

Claude Code and Codex alerts

Choose Install and Elissa tells you when a session finishes or needs your permission.

• It keeps your settings and saves a backup (*.elissa-backup).
• In Codex, approve it once with /hooks.
• It only sends the kind of alert and the project's name.

Using the computer

Windows and macOS. Let her see and use other windows: read the window you're in, click, type and press keys, always with your approval. Password fields are never read. On macOS, allow Elissa in Accessibility when she asks.

Folders she can use

Where she can find, read and write files. Your home folder by default; Add folder for more. Files you drop on her work from anywhere.

Memory

New conversation starts over. Each conversation is summarized into notes she can recall later. The bottom of the panel shows your version and Update now when there's a new one.