Use AI writing tools

You can write and organize a local library without enabling AI. Built-in AI writing tools use the provider and model you configure on this device. They are separate from MCP Agent Access, which lets another app read your library and, with separate permission, change documents, organization, and notes.

Cloud providers need an internet connection and your own API credentials. Provider usage can have separate charges and account requirements. A local model server can run on the desktop without a cloud provider once it is installed and ready.

1. Open Settings → AI → Models.
2. Turn on Enable AI features and choose a Provider.
3. For a cloud provider, enter your API key. For a local provider such as Ollama or LM Studio, start its model server on this computer and check the API base address.
4. Choose a Model from the available list, or enter the exact model ID. Use Refresh models if the list is missing or out of date.
5. Choose Save settings.

You can request Rewrite, Make shorter, Make longer, Fix grammar, or Translate…. Review suggestions for accuracy before accepting them.

1. Select text in your document.
2. Choose Ask AI in the formatting toolbar and select an action.
3. Review the generated suggestion. Choose Accept to apply it, Discard to leave your text unchanged, or Regenerate to try again.

Enable Suggest text while writing in Models settings, then save. Pause at the end of a sentence to request a suggestion. Press Tab to accept it or Escape to dismiss it.

Inline model can use a different model for quicker responses. Leave it blank to reuse the main model. You can turn inline suggestions off while keeping other AI actions enabled.

AI requests can include your selected writing and surrounding document context. With a cloud provider, that content is sent to the configured model service; review that provider’s data practices before using private writing.

Configuration stays on this device. Desktop API keys use the operating system’s protected storage when available and are not written into library files. A custom API base sends requests to that address, so use an endpoint you trust.

Local model servers are desktop-only. Mobile AI uses your own cloud credentials. Turning off Enable AI features removes AI assistance and suggestions; save the change.

Confirm that AI is enabled, a model is selected, and settings were saved. If model discovery fails, check the API base and credentials, then try Refresh models or enter a model ID manually.

For a local model, check that its server is running and the model is available. For cloud models, check your connection, provider access, and usage limits. A slower model can make inline suggestions impractical; try a faster inline model.

If the problem continues, send support the provider name, model ID, app version, and error message. Never include API keys or private document text.