How to enable BYOK AI and switch providers
NexoPad AI works in BYOK mode: you choose the provider, save your key, and control the model.
The extension does not lock you into a single provider. You can store several keys, activate one or another depending on the case, and choose a different model for each service.
The approach remains BYOK: the key comes from your own account and is used to call the chosen provider, not to send it to a NexoPad backend.
How to activate AI
Open AI settings from the side panel or popup
The current version lets you manage multiple providers from one place and activate the one you want to use at that moment.
Save one or more API keys
You can keep several keys at once and switch between them without redoing the entire setup each time. The interface tries to validate the key before it becomes the active configuration.
Activate the provider and choose the model
You can use the suggested default model or type a custom name if your account supports another compatible model.
Adjust persona, options, and target language if needed
Besides the provider itself, you can now also define a persona or system prompt and other settings tied to the AI experience.
Go back to the panel and test a real text action
That confirms in a real scenario that your provider, key, and model are working as expected.
Providers and default models detected in the extension
| Provider | Default model | Useful note |
|---|---|---|
| Groq | llama-3.3-70b-versatile | Often a fast option to start with your own key. |
| Mistral AI | mistral-small-latest | Additional provider supported directly by the extension. |
| Cohere | command-r-plus-08-2024 | Available as a native option in the provider selector. |
| OpenRouter | openai/gpt-oss-120b:free | Presented by NexoPad as a useful path for accessible or free catalog models. |
| OpenAI | gpt-4o-mini | You can switch to another compatible model if your account supports it. |
| Anthropic | claude-haiku-4-5-20251001 | Designed for fast answers with contained cost. |
| Google Gemini | gemini-2.0-flash | A lighter option for short tasks. |
| DeepSeek | deepseek-chat | Can also switch to other compatible models from the provider. |
| xAI (Grok) | grok-3-mini | Supports custom model names if your account allows it. |
| Together AI | meta-llama/Llama-3.3-70B-Instruct-Turbo | Another provider directly supported in the current selector. |
Privacy, errors, and availability
The extension is built with a BYOK approach: your API key is stored to call the chosen provider directly, not to send it to NexoPad-owned servers.
If you see a provider unavailable error, check four things first: whether the key is still active, whether you have quota or credit when applicable, whether the model exists for your account, and whether the provider is available in your region.
The availability of free or low-cost models can change over time depending on the provider policy. The important point is that NexoPad exposes the selector, but the real catalog depends on your account and on the external service.