Mouse-1

Mouse-1 is a decision model API from Mouse: you send an agent's current state and a list of options, and it returns a probability for each, so an agent harness can make small routine choices quickly without calling a frontier model.

Why it matters

For developers building agents, it offloads the many small yes-or-no and pick-one decisions in a task to a cheap, fast classifier, leaving the expensive model for reasoning and code.

Highlights

• Takes the agent's state and options and returns probabilities; one request can answer up to 128 questions in about 100 ms, according to Mouse.
• Example uses on the launch page include image moderation, document routing and detecting what state a browser screen is in.
• Only picks from the options you give it; Mouse reports planted instructions changed 0% of decisions in its tests.unverified
• Mouse states zero data retention: requests are scored and dropped, not stored or used for training.
• Mouse reports 89.1% average decision accuracy on unfamiliar tools and repositories in its own benchmark.unverified

Taken from the project's own README, docs or launch page. Claims about speed or accuracy are the maker's unless stated otherwise.

Worth knowing: Accuracy, speed and injection-resistance figures are Mouse's own benchmarks. Mouse also sells a separate mobile coding-agent app (Pro $60/mo, Max $200/mo) that is not part of this API.

Quick facts

Links

• Official site: https://mouse.dev/api
• Release notes: https://mouse.dev/blog
• Docs: https://mouse.dev/docs

Related ranking

See how established tools compare in Best AI Coding Tools for Developers. New releases aren't scored until we review them against the rubric.

What we couldn't verify

• highlights[2] (maker's test)
• highlights[4] (maker's benchmark)

Checked 2026-10-21 from the model's own pages. Prices and star counts change, so confirm on the official site. Listing here is not a ranking or an endorsement.