Anyone turning the AI on
AI providers and keys
- Where a key livesThe one page to read before pasting an API key: the three places a text or image backend can be set (the device, the campaign, the server), which one wins, who can see a key, and what needs tool calling.Desktop appAndroid appServer
- OpenAI and OpenAI-compatible keysConnecting OpenAI in the apps and on a server, the model defaults, how one key covers narration and pictures, the 128K context, the strict-vendor handling for reasoning models, and using any other OpenAI-compatible endpoint.Desktop appAndroid appServer
- OpenRouter and other routersUsing OpenRouter to reach Claude, Gemini, GPT or an open model with one key: the settings, picking a model with tool calling, what the server sends to a router, and cost and privacy notes.Desktop appAndroid appServer
- Local models: llama.cpp, Ollama, LM Studio, vLLMRunning the Dungeon Master on your own GPU: the recommended model and its exact settings, the llama.cpp command, the Ollama Modelfile, LM Studio and vLLM, thinking mode and the reasoning budget, why presence penalty is pinned, and what to expect from a turn.Desktop appServer
- Bring your own agentLetting Claude Code, Codex, opencode or Grok Build narrate on the plan you already pay for: what the app checks, how the program is boxed in, choosing story and bookkeeping models, testing a turn, pictures, and the builds where it is not available.Desktop appServer
- Image providersWhere pictures come from: an OpenAI key, ComfyUI on your GPU, the FLUX worker, or an agent's own image tool. What each paints, what it costs, the desktop app's guided ComfyUI install, and what never needs an image backend.Desktop appAndroid appServer
- What leaves your machineA plain accounting of where data goes in each setup: nothing with a human Dungeon Master or a local model, the prompt with a remote key, the tunnel and the broker when sharing, voice relays, the world registry, and what the project itself never sees.Desktop appAndroid appServer