AI setup
Deskless offers free on-device coding help on supported devices and cloud AI with your own provider API key. Cloud Agent Mode is available on both iOS and Android.
New in 3.0: Grok can use Sign in with Grok, and Agent Mode 2.0 works with Claude, OpenAI, Gemini, Grok and OpenRouter. Deskless 3.0 is available for iPhone and iPad; Android 3.0 is in review on Google Play.
Free on-device help
- iPhone and iPad: use Apple Intelligence for quick coding help on supported devices.
- Android: use Gemini Nano for supported on-device coding actions when the device and model are ready.
On-device help does not require Deskless Pro or a provider API key, and prompts are processed on your device. Availability depends on device support, system settings, and model readiness; installing Deskless alone does not guarantee access. Check the AI options available in the app.
Quick on-device assistance handles selected-code actions. The separate Offline Agent adds reviewed local project edits on supported devices: Beta on iOS, Experimental on Android 3.0. Cloud Agent Mode requires Deskless Pro and a cloud provider API key on both platforms.
Cloud AI with Pro
For cloud AI actions and Agent Mode, choose your provider and model, connect your own API key, and pay the provider directly for usage. Requests go from your device to the configured service without a Deskless relay. OpenRouter routes requests onward to the provider serving your selected model.
Supported providers
| Provider | Create a key |
|---|---|
| OpenAI | platform.openai.com/api-keys |
| Anthropic | console.anthropic.com |
| aistudio.google.com/apikey | |
| xAI | console.x.ai |
| OpenRouter | openrouter.ai/settings/keys |
Available model names can change. Deskless shows the models supported by the current app and provider configuration.
Grok can use an xAI API key or, in Deskless 3.0, Sign in with Grok. Sign-in opens xAI’s consent page with a device code; eligible plans and available models are controlled by xAI. Android 2.1 offers the API key only.
Add a provider key
- Open Settings → AI Providers.
- Choose a provider.
- Paste the API key created in that provider’s console.
- Select a default model if prompted.
- Save and run a small Explain request to confirm the connection.
Keys are stored using the secure credential storage provided by your operating system. Deskless does not send them to its own servers.
Before a first content request to a provider/destination, Deskless asks permission to send the relevant prompt, code, files or diffs. Declining prevents that request. For OpenRouter, consider both OpenRouter and the downstream model provider when reviewing the content permission and service terms.
One-shot AI actions
Select code, open the AI actions menu, then choose Explain, Improve, Add Comments, or Find Bugs. Actions that can edit code show the result for review before you insert or replace anything.
Streaming replies
Replies appear as they arrive. Wait for completion before applying a quick action that changes code; interrupted, canceled or failed partial output is not a finished edit. You can stop a request and retry after checking the connection, provider key, quota or reported error. Streaming does not remove provider usage charges or rate limits.
Agent Mode
Use Agent Mode for multi-file, multi-step work. It can inspect the repo, build a plan, target relevant files and symbols, propose patches, verify risky edits, and present a grouped review. See Agent Mode for the complete workflow.
Privacy and cost
- On-device AI processes prompts locally. Cloud requests go to the configured service; OpenRouter also forwards them to the selected model provider.
- The provider’s terms, retention settings, rate limits, and pricing apply.
- Deskless does not sell AI credits or add a usage markup.
- Send only code and context you are authorized to share.
- Always review and test generated changes.