Author Studio

Models & privacy

Your words.
Your connection.

The desktop app connects directly to your chosen AI provider or server. There is no Author Studio account, analytics, or tracking.

This guide covers desktop connections. The plugin uses your host's models, tools, storage, and privacy policies instead; it has no direct model API integration.

Claude and GPT

Create an API key at Anthropic or OpenAI. Add it to a connection in Settings, choose a model, and use Test connection.

Cloud services bill API usage separately. Check the provider's model pricing and data-handling policy before sending a manuscript. Writing and relevant project context are sent to that provider; optional reviews and confidence-assessment follow-ups use additional tokens.

Local and network models

Use Ollama, LM Studio, llama.cpp, or another server that speaks the OpenAI API. Start the server, select the connection type, check the address, and choose Find models. An optional API key supports authenticated servers and hosted OpenAI-compatible services.

A server can run on this computer or another trusted computer on your network. For example, an Ollama server may be at 192.168.1.20:11434.

  • Ollama: to allow network connections, set OLLAMA_HOST=0.0.0.0 on the server computer and restart Ollama.
  • LM Studio: turn on Serve on Local Network in the Developer tab.
  • macOS: allow Author Studio to find devices on your local network when prompted. Change this later under Privacy & Security → Local Network.

Only expose a model server on a trusted network, with appropriate firewall and authentication controls. A network connection sends your writing to that computer. A hosted compatible endpoint is still a cloud service, not a local model.

Set a realistic context window

Under Advanced → Context window, enter the capacity your server is configured to use. The app also asks the server for its current capacity and uses the smaller value. If a prompt is rejected as too large, it retries with less background material and remembers the smaller size for that model on that server.

Small local models tend to write shorter, simpler chapters and need more human review. A larger context window does not by itself guarantee better continuity or prose.

Apple Intelligence

Requires an Apple silicon Mac, macOS 26 or later, and Apple Intelligence enabled in System Settings → Apple Intelligence & Siri. The app reports when requirements are missing.

Generation runs on the Mac with no per-request API charge. Apple's small on-device model reads roughly 3,000 words at once, so the app supplies shorter instructions and less story context. It is better suited to notes and short pieces than long, complex chapters. Expect more review.

Allow mature themes is on by default and selects Apple's less restrictive content filter for fictional crime, conflict, and darker material. Disable it in the connection settings to use Apple's standard filter.

Where your data lives

Desktop projects are stored on your computer. Open Help → Show Projects Folder to find them.

  • macOS: ~/Library/Application Support/Author Studio
  • Windows: %APPDATA%\Author Studio
  • Linux: ~/.config/Author Studio

API keys use the operating system's protected storage: macOS Keychain, Windows data protection, or a Linux keyring. Without GNOME Keyring, KWallet, or an equivalent Linux keyring, keys receive only basic protection; Settings warns you.

A saved key is sent only to the service or server it was entered for. Changing a server to a different address requires entering its key again. Writing goes to your selected model connection; local servers and Apple Intelligence keep generation on your own computers.

App update checks contact GitHub, not your AI provider, and do not send project text or API keys. This documentation site has no analytics scripts, third-party fonts, or client-side trackers. Opening a download picker contacts GitHub's API to find the latest stable installers. The site is hosted by GitHub Pages, whose own service policies apply.

Backups, exports, and portability

Use History → Back up project to save a complete project copy, and File → Open Backup or State File to restore it. Keep backups somewhere you control and protect them like your manuscript.

Readable manuscript and planning exports are useful for sharing but do not replace backups. Plugin state files carry the story bible and progress, not the entire manuscript. See export options for the distinction.