Getting started
DocsGetting started
Start hereGetting started
Opera is a privacy-first AI assistant. You can run a model on the device itself, connect your own Ollama server, or use any OpenAI-compatible endpoint. There is no account, and nothing leaves your device unless you choose an online model, cloud sync or web search.
What you need
- An Android phone or tablet, a desktop computer, or a modern browser.
- For on-device models: enough free storage and memory for the model file you want to run.
- For Ollama or another local server: the server running on a machine Opera can reach.
- For an online provider: an API key from that provider.
Install Opera
| Platform | Where to start |
|---|---|
| Android | Join the beta or build a dev client |
| Desktop | Download a release or build from source |
| Browser | Open the web build from the Opera site, or run it locally with npm run web |
| Your own server | Install Sampler Hub for sync and shared models |
First run
- Choose a provider. Pick a built-in model, an on-device model, Ollama, or an OpenAI-compatible endpoint.
- Download a model if needed. On-device models are stored locally and survive app updates. You can remove them later from settings.
- Grant only the permissions you want. Camera, microphone, contacts, calendar and screen access are optional and can be toggled individually.
- Send a message. Ask a question, attach a PDF, or use the microphone.
If you want Opera to replace Google Assistant, set it as the default assistant. This is optional and can be reversed at any time.
Keep it private
Opera’s defaults are local-first:
- Conversations are stored in the app’s private SQLite database.
- Incognito conversations are never written to history.
- Cloud providers only receive the messages you send while they are selected.
- Cloud backups are encrypted with a PIN; Sampler Hub sync is sealed with the profile key.
- Shared conversations use encrypted PrivateBin links, with the key kept in the URL fragment.
Read Privacy for the full data map.