Choosing Wisper's brain

Wisper’s brain is the AI she thinks with. WispOS doesn’t come with one: you bring your own, and you can change it whenever you like in Settings → Wisper & voice, during setup, or from the brain button in the app’s header.

The two kinds of brain

Agents that can act

Claude Code and Codex (ChatGPT) are command-line agents. With either, Wisper works inside your HQ: she can read and write files, run your HQ’s scripts, start work on your projects and hand it to teams. This is the full WispOS.

Claude Code is what Wisper is built around and what we recommend. A few things currently need it, including shared team agents.

Models that talk

Any OpenAI-compatible model can be her brain for conversation:

• On this computer: Ollama, LM Studio, or a llama.cpp server.
• Online services: OpenAI, OpenRouter, Google Gemini (with an API key), Groq, Mistral, DeepSeek, xAI, Together, or any compatible address you type in.

These can answer and talk things through, but they can’t run HQ’s scripts or touch files, so they can’t do the work themselves or dispatch it. The choice list says so next to each one.

An API key you enter is kept by the app’s own back end and never shown again in the window.

Your plan, your limits

WispOS uses your own account with the AI provider. The Usage screen shows how much of your plan’s limits you’ve used and when they reset, as reported by the provider’s own tools on this computer.

Privacy

What you say to Wisper, and what she reads to answer you, goes to the brain you chose. A local model keeps it on your computer.