Fast API that scores agent decisions in ~100ms, with choice, yes/no, and rubric scoring primitives.
Mouse-1 API homepage
1 / 3Overview
What is Mouse-1?
A decision model for your agent harness.
Mouse-1 is a decision model built for AI agent harnesses. Instead of generating text token by token, it reads your agent's state and scores the options directly, returning calibrated probabilities in about 100ms.
A single request can evaluate up to 128 typed questions in parallel, each using one of three primitives: choice (pick from up to 255 named options), noul (yes/no condition), or score (2–10 ordered rubric levels). If confidence is low, the model can return an abstain instead of guessing.
By default Mouse stores none of your data and never trains on it. Because it only scores the options you provide, it is naturally resistant to prompt injection — a useful property for automated agent pipelines.
Pricing is usage-based: the API starts at $0.10 per 1M input tokens with output free, and new signups get $5 in starter credits after verification. Pro ($60/mo) and Max ($200/mo) plans both include 2-day trials.
Platforms and languages
Mouse-1 Availability
Platforms
Languages
Capabilities
Mouse-1 Key Features
Non-autoregressive decision model
Scores options directly instead of generating text; evaluates up to 128 typed questions in parallel per request and returns calibrated probabilities in about 100ms.
Three decision primitives
choice picks from up to 255 named options, noul answers binary conditions, and score rates 2–10 ordered rubric levels, each returning a schema-valid distribution.
Zero data retention by default
Your data is not stored and is never used for training, making it a fit for pipelines handling sensitive agent state.
Prompt-injection resistant
The model only scores the options you provide and never free-generates text, so injected prompts cannot expand its output surface.
Calibrated abstain
When confidence is low, Mouse-1 returns abstain instead of guessing, so agents can fall back to safe default behavior.
Best for
Who uses Mouse-1?
Developers and engineering teams building AI agents or agent harnesses
Agent micro-decisions are slow and expensive with frontier LLMs; need faster, cheaper scored decisions
Plans and access
Mouse-1 Pricing
Freemium
API from $0.10/1M input tokens (output free); $5 starter credits; Pro $60/mo; Max $200/mo.
Common questions
Mouse-1 FAQs
How is Mouse-1 different from a regular LLM?
It does not generate text at all. It scores the options you give it and returns calibrated probabilities in about 100ms, which makes small agent decisions much faster and cheaper than calling a frontier LLM.
Is there a free tier or free credits?
Yes. Verifying your account after signup gives you $5 in starter credits, and the API itself is pay-as-you-go starting at $0.10 per 1M input tokens with free output. Pro ($60/mo) and Max ($200/mo) include 2-day trials.
Does Mouse keep my data?
By default, no. Mouse does not store your data and never uses it for training.
How do I integrate Mouse-1?
Via the REST API at api.mouse.dev, with TypeScript and Python SDKs plus an interactive Playground. The service runs in US-East.
How does it handle prompt injection?
It only scores the options you provide and never generates free text, so injected prompts cannot change what it outputs. It also returns a calibrated abstain when confidence is low.
Reviews
See what the community thinks and share your experience.
—
Based on 0 ratings
Rating distribution
Community reviews
No reviews
No written reviews yet.
