Mouse-1

Mouse-1 is a decision-model API from Mouse that its website calls "a decision model for your agent harness." You send it the current state and a typed question, and it returns a probability for each option, which developers can use to route tasks, check results or gate whether a step goes ahead. It only returns probabilities and doesn't run commands or touch files itself. This listing covers only the API, not the phone remote-control features in Mouse's main product.

Who it's for:Developers building AI agent systems who need a dedicated decision step. Readers who don't write code won't have a use for it.

Before you start: You need a console account and an API key, and you can connect over HTTP or with the official JavaScript or Python clients, without installing Mouse's main product. Check the website for current prices: it lists USD 0.10 per million input tokens with free output, and the docs mention USD 5 in starting credit. You prepay, and automatic top-ups only happen if you turn them on, with the terms limiting automatic card charges to once a day. The state, questions and optional images you send go to api.mouse.dev in the US East region; the company says it keeps no data, doesn't train on it and doesn't log requests, which is its own promise and hasn't been independently audited. The docs include hook examples that send agent instructions and commands to the cloud for a decision, and whatever happens after you integrate it is up to your own code. A high or low probability isn't a permission or a safety guarantee. The source code isn't public.

Based on the official site, documentation and terms of service. Not tested by AI Inspiration World.