On-Device AI, Bring Your Own Key or Keyda AI: Which Route Should You Use?

AI is not one-size-fits-all. Sometimes you want maximum privacy. Sometimes you want the model quality or speed of a cloud provider you already use. Sometimes you just want the app to handle the setup for you.

Keyda is built around that reality. Instead of forcing every user into one AI path, Keyda supports three routes: on-device AI, bring your own key and Keyda AI. Each route has a different balance of privacy, convenience and capability.

This guide explains how to think about them.

The short version

Use on-device AI when privacy, offline access and local processing matter most, and when the task is supported by the local model available on your device.

Use bring your own key when you already have an AI provider account and want requests to use your own key.

Use Keyda AI when you want managed cloud AI access through Keyda's trial, Lite or Pro plans without setting up provider keys yourself.

The best option can change by task. A private note may call for local AI. A complex rewrite may be better with a cloud model. A casual smart reply may be easiest through managed Keyda AI.

Route 1: On-device AI

On-device AI runs supported model work locally on the phone. In Keyda's iOS app, local models use Apple's device capabilities through MLX; Android uses its own local model path. Availability and performance depend on the device and model.

The main benefits are:

• Privacy: supported requests can stay on the device.
• Offline access: local AI can work without a cloud request when the model is available.
• No provider key required: you do not need to bring an external AI account for local tasks.

The tradeoff is that local models can be more limited than large cloud models. They may be slower on older devices, and some tasks may still be better handled by a cloud route.

Choose on-device AI for sensitive drafts, private notes, offline work and lightweight rewriting where local quality is enough.

Route 2: Bring your own key

Bring your own key, often called BYOK, lets you connect your own AI provider key. Keyda can then route supported cloud requests using your provider setup rather than Keyda's managed AI pool.

This is useful when:

• You already pay for or manage an AI provider account.
• You want control over provider choice.
• You prefer your own billing relationship with the model provider.
• You want cloud-model capability without relying on Keyda's managed AI quota.

BYOK is a good fit for power users and teams who already know which provider they want. It offers flexibility, but it does require setup and responsible key management.

Route 3: Keyda AI

Keyda AI is the managed route. Instead of adding your own provider key, eligible users can send AI requests through Keyda's backend service. This is the convenient option for people who want AI writing help without configuring external provider credentials.

Keyda AI is designed for:

• Trial users testing AI keyboard workflows.
• Lite users who want managed access for everyday writing.
• Pro users who need more frequent AI help.
• Users who do not want to manage API keys.

The tradeoff is that managed access is tied to plan limits and availability. For many everyday tasks, that simplicity is exactly the point.

How to choose the right route

Think about four questions:

1. How sensitive is the text?

If the content is private, local processing is often the first route to consider when supported. If the content is casual, a cloud route may be fine.

2. Do you already have an AI provider key?

If yes, BYOK can make sense. If no, Keyda AI avoids that setup.

3. Do you need offline support?

On-device AI is the route designed for offline-capable workflows when the model is downloaded and available.

4. How much setup do you want?

On-device models and BYOK can require more setup. Keyda AI is the simplest path for users who want to start quickly.

A practical comparison

Use case	Best starting route
Private note rewrite	On-device AI
Casual smart reply	Keyda AI
Power user with provider account	Bring your own key
Offline draft cleanup	On-device AI
Quick setup for new users	Keyda AI
Provider-specific workflow	Bring your own key

This is not a strict rule. It is a starting point. Keyda's advantage is that users can move between routes depending on the job.

Why Keyda supports all three

Most AI apps make a single assumption: either everything is cloud-based, everything is local or everything runs through one managed service. Real users are messier than that.

Someone might want local AI for private notes, their own provider key for advanced work and managed Keyda AI for quick everyday replies. Those needs can all exist in the same week.

Because Keyda is a keyboard, flexibility matters even more. The keyboard appears inside many apps and contexts. A one-route AI model would be too rigid for that surface.

The privacy mindset

The safest habit is to match the route to the content. Ask:

• Would I be comfortable sending this text to a cloud provider?
• Do I need the best available cloud model for this task?
• Is convenience more important than local processing here?
• Can a local model handle this well enough?

Keyda gives you options so the answer does not have to be the same every time.

Final recommendation

If you are new to Keyda, start with Keyda AI to understand the workflow. Then explore on-device AI for private or offline tasks. If you already use an AI provider, add your own key and compare results.

The point is not to pick one route forever. The point is to have AI writing help that adapts to the moment: local when privacy matters, BYOK when control matters and Keyda AI when convenience matters.

Keyda's AI routes are not competing philosophies. They are different tools for different typing moments.