FAQ

Lokei is a private AI assistant that lives in your Mac's menu bar. On supported Macs it can use Apple Intelligence for on-device responses, and it also supports local Ollama models plus optional OpenAI-compatible providers you configure.

You need a Mac running macOS 13 Ventura or later. Apple Intelligence support depends on your Mac model, macOS version, and Apple Intelligence availability. If you choose the Ollama backend, you also need Ollama installed and running.

Lokei works on any Mac running macOS 13 Ventura or later, including both Intel and Apple Silicon Macs. Apple Intelligence is only offered on Macs that support it; other Macs keep the Ollama-first setup flow.

Install Lokei from the Mac App Store and follow onboarding. Supported Macs can start with Apple Intelligence. If you choose Ollama, Lokei shows the Ollama download link and model setup steps inline, then lets you manage models from Settings → AI & Models.

Lokei is built around local-first privacy. With Apple Intelligence or local Ollama, processing stays on your Mac. If you configure a remote provider, prompts sent to that provider leave your Mac and are governed by that provider's terms — but Lokei itself still does not collect them. Lokei does not operate chat servers, does not require an account, and does not include analytics or tracking.

Lokei collects zero data on Lokei servers: no conversations, no usage stats, no analytics, nothing. Your chat history, settings, and project context are stored locally on your device. If you enable iCloud sync, that data syncs through your Apple iCloud account — not through Lokei servers.

No Lokei account is required. App Store distribution is handled by Apple. A separate provider account is only relevant if you choose to configure a remote provider yourself.

Yes, with local or on-device backends. Apple Intelligence availability is handled by macOS, and Ollama works offline after Ollama and your models are installed. Remote providers require internet access.

If you use Apple Intelligence, macOS manages the model. If you use Ollama, it depends on your Mac's RAM: smaller 1–3B models are best for 8 GB Macs, while 16 GB+ Macs can usually run larger 7B models. Lokei's Model Library shows RAM requirements for every Ollama model.

No. Apple Intelligence needs no model download inside Lokei. For Ollama, Lokei includes a built-in Model Library where you can browse and download models with a single click. Models are rated by RAM requirement so you know what should run well on your Mac.

Small models (1–3B parameters) are typically 1–2 GB. Medium models (7B) are around 4–5 GB. Lokei's Model Library shows the exact download size for each model before you download it. Models are stored in Ollama's directory, not inside the Lokei app itself.

Yes. Lokei automatically detects all models installed through Ollama and shows them in the model picker. If you've been using Ollama before Lokei, all your existing models are immediately available.

Advisor Mode lets Lokei escalate from a fast model to a stronger model only when needed — for deeper reasoning, confidence checks, complex coding decisions, or high-stakes tradeoffs. Use a fast model for everyday work and escalate when the moment calls for it. This feature is in development.

Adaptive Response Mode is a planned feature where Lokei privately detects the kind of help you seem to need and adjusts the response style — whether that's a quick fix, a calm explanation, deep analysis, or a focused next step. All detection happens on your device with no profile built in the cloud. This feature is in development.

An iOS companion app is in development. It's being built around the same privacy posture: Apple Intelligence on eligible devices, local-network AI when you use your Mac as the server, and no Lokei cloud in the middle. It is not on the App Store yet. Availability, features, and requirements may change before release.

The planned approach is for your iPhone to connect to Ollama or LM Studio running on your Mac over your local Wi-Fi network. This keeps your AI processing on hardware you control without routing through external servers. Details may change before release.

Lokei is planned around ownership, not subscriptions. Final pricing is shown on the App Store. There is no Lokei subscription planned.

Lokei itself has no subscription. Apple Intelligence and Ollama do not add per-message API fees. If you configure a remote provider, that provider may have its own pricing or usage limits. Optional remote providers may have their own pricing.

Apple's App Store policies determine use across Macs associated with your Apple ID, including any applicable Family Sharing support.

This only applies if you choose the Ollama backend. Open your Applications folder and launch the Ollama app, or run ollama serve in Terminal. Once Ollama is running, click "Try Again" in Lokei. If you haven't installed Ollama yet, download it from ollama.com.

For Ollama, switch to a smaller model — 3B parameter models are significantly faster than 7B models, especially on 8 GB RAM Macs. Long conversations can slow down over time, so use "Summarize & Compress" in the chat or start a new chat to speed things back up. Remote provider speed depends on the provider and your connection.

Email us at hello@getlokei.com. We're a small team and we read every message.