AI features and your own API keys
Arino One includes AI features that help agents work faster: suggested replies, automatic conversation summaries, inbound translation, and synthetic voice for greetings and voicemail. Text AI runs on Lovable AI (managed) by default — it works out of the box with no key — or you can point it at your own OpenAI or Anthropic key. Either way AI usage is metered, and Arino adds no markup.
AI is optional, but it is not free. Text AI has a master on/off switch in Settings → Build with AI. With it on, Arino One uses Lovable AI by default and usage is metered in Lovable AI credits; you can switch the provider to your own OpenAI or Anthropic key at any time.
What AI features are available
AI-assisted replies suggest a next response while an agent is reading a conversation. The agent sees a draft, can edit it, and sends it as a normal message. It is a drafting aid, not an auto-send: the agent remains in control of every outbound message.
Conversation summarisation generates a short recap of a long conversation on demand. This is useful when an agent takes over a ticket, when a manager reviews escalated threads, or when you need a handover note for a channel that has been running for hours.
Translation translates inbound messages into the agent's language, and can translate outbound replies back into the contact's language. It is used on channels where contacts write in languages your team does not cover, and it is enabled per channel rather than globally.
ElevenLabs voice turns written greetings, hold messages and voicemail prompts into spoken audio. It is a separate, optional integration from the text-AI provider choice, and it is used by the voice routing system so callers hear a natural voice rather than a recorded clip or system beep. See voice and WebRTC calling for how voice routing, recording and voicemail work.
Choosing your AI provider
Text AI providers are chosen in Settings → Build with AI, where the master switch also lives:
| Provider | Key | Billing | What it enables |
|---|---|---|---|
| Lovable AI (managed) — default | None required | Metered in Lovable AI credits, billed by Lovable | AI-assisted replies, summarisation, translation |
| Anthropic (Claude) | Bring your own Claude API key | Billed by Anthropic | AI-assisted replies, summarisation, translation |
| OpenAI | Bring your own OpenAI API key | Billed by OpenAI | AI-assisted replies, summarisation, translation |
| ElevenLabs (voice, separate) | Bring your own ElevenLabs API key | Billed by ElevenLabs | Synthetic voice greetings and voicemail prompts |
Lovable AI needs no setup: turn text AI on and it works. Bring your own key only when you want usage billed to your own provider account or you need a specific model. ElevenLabs is not part of the text-provider choice — it is configured with the voice/secrets setup.
How to add your own key
- Open your instance console and go to Settings → Build with AI.
- Switch the provider from Lovable AI to OpenAI or Anthropic.
- Paste the API key from the provider's dashboard. For OpenAI, this is a project or user API key from platform.openai.com. For Anthropic, it is a key from console.anthropic.com.
- Save. The key is encrypted at rest and is only readable by the server-side functions that call the provider.
For ElevenLabs, add the key from the API keys section of your ElevenLabs account in the voice/secrets setup on your instance. Arino One does not generate or sell API keys on your behalf.
Security and privacy
API keys are stored in your instance's server-side secrets store and are never exposed to the browser or the mobile app. Agents see the output of AI features — a suggested reply, a summary, a translation — but they never see the key itself. Arino One does not train models on your data or share conversation content with any provider other than the one you explicitly configured.
Because usage is metered, it is always billed to someone: Lovable for the managed default, or your own provider account when you bring a key. Arino One does not add a markup on AI tokens or voice minutes. If you turn the master switch off, AI features stop immediately and no further provider calls are made.
Where AI features appear
- Inbox: suggested replies appear below the composer when text AI is on.
- Conversation panel: a summary button appears on long threads when summarisation is enabled.
- Channel settings: translation can be toggled per channel for inbound, outbound or both directions.
- Voice routing: ElevenLabs voice is selected as a voice prompt source in Settings → Voice once an ElevenLabs key is added.
If text AI is switched off, the corresponding controls are hidden or disabled. This keeps the UI clean for teams that do not want AI features.
Costs and billing
AI is metered, never free. By default it is metered in Lovable AI credits and billed by Lovable; when credits are exhausted, AI pauses until Lovable is topped up. If you bring your own OpenAI or Anthropic key, that provider meters and bills you per token instead. ElevenLabs is billed per character of generated audio on your own ElevenLabs account. Arino adds no markup on any of it.
Arino's $50/month connection fee covers instance management, seats and messaging connection — it does not include AI usage. To control spend, you can set model limits and feature toggles in Settings → Build with AI, or turn the master switch off entirely. See billing and plans for what is included in the Arino connection fee.
Troubleshooting
If AI features stop returning suggestions, check Settings → Build with AI first. On the managed default, the most common cause is exhausted Lovable AI credits — top up and AI resumes. On a bring-your-own key, the key may have been revoked in the provider dashboard, hit a billing limit, or been removed from the instance; re-adding or regenerating the key fixes most cases. If the key is valid but a feature still fails, the provider may be returning an error; the server logs surface the provider status without exposing the key.
Frequently asked
Do I have to use AI features?
No. Text AI has a master on/off switch in Settings → Build with AI, and you can leave it off. When it is on, it runs on Lovable AI (managed) by default, so it works immediately without you adding any key — usage is metered in Lovable AI credits.
Which AI providers can I use?
For text AI: Lovable AI (managed) is the default and needs no key, or you can bring your own OpenAI or Anthropic (Claude) key. ElevenLabs is a separate, optional integration for synthetic voice greetings and voicemail prompts.
Where are my API keys stored?
In your instance's server-side secrets store. Keys are never sent to the browser, never exposed in the UI, and never logged. Arino One reads them only when making provider calls on the server.
Who pays for AI usage?
By default you spend Lovable AI credits, billed by Lovable — AI pauses until you top up when credits run out. If you bring your own OpenAI, Anthropic or ElevenLabs key, that provider bills you directly. Arino adds no markup either way.
Can I use the same AI key across multiple instances?
Yes, but you manage it per instance. Each instance has its own Settings → Build with AI page, so you add the key where you want it used. This keeps billing and data separation clean between instances.
Apply this to your own deployment
This guide describes decisions we make on live instances. Tell us your channels, systems and region and we will map it to an architecture outline, a provisioning plan and an indicative commercial model — usually within one business day.