Mouse-1
Claim this toolRoute, gate, and verify agent decisions in 100 ms, up to 65x cheaper
From $60/mo, billed monthly, free trial Visit Website Paid Visit Website
Tracked since 2026
The Bottom Line
- Price
- From $60/mo, billed monthly, free trial. Compare plans
Key facts
- Mouse-1 is a decision model that handles small, frequent decisions in agent harnesses, freeing frontier models for complex reasoning.
- It delivers decisions in ~100ms with 89.1% accuracy on unfamiliar tools and 0% prompt injection success rate.
- Priced at a fraction of frontier models, it uses a pay-per-input-token model with no data retention.
Pros
- Extremely fast (~100ms) decision latency
- Up to 65× cheaper than frontier models for decision tasks
- High accuracy on unfamiliar tools and strong resistance to planted instructions
Cons
- Limited to decision-making; cannot replace frontier models for reasoning or code generation
- Requires integration into an existing agent harness, not a standalone tool
What is Mouse-1?
Mouse-1 is a specialized decision model designed to handle the hundreds of small, rapid decisions that AI agents make during every task. It plugs directly into an agent harness, taking the agent's current state and a set of options, and returns a probability for each option. This allows the main frontier model to focus on reasoning and code generation while Mouse-1 efficiently routes, gates, and verifies decisions in about 100 milliseconds. The model is up to 65 times cheaper than frontier models and delivers high accuracy even on unfamiliar tools and repositories.
Mouse-1 excels in use cases like content moderation, document routing, browser state detection, and safety verification. It is resistant to prompt injection, with a measured 0% decision flip rate in tests. The API operates with zero data retention: every request is scored and immediately discarded. Pricing is pay-per-use based on input tokens, with no cost for output tokens. A 2-day trial is available for the Pro and Max subscription plans, which include hosted cloud agents, GitHub PRs, and higher usage limits.
Data compiled by Toolradar Pricing checked · How we evaluate
Source mouse.dev Key Features
- Takes agent state and options to return a probability for each option
- Handles up to 128 decisions per request in about 100ms
- Zero data retention: requests are scored and immediately discarded
- Resistant to prompt injection with 0% decision flip rate in tests
- Supports use cases like content moderation, document routing, and browser state detection
- Integrates directly into any agent harness via a simple API
Pricing Plans
Free TrialPricing checked Oct 23, 2026. Source: mouse.dev/pricing, read Oct 23, 2026
| Plan | Price | Details |
|---|---|---|
| Pro | $60 / mo |
+1 more Show less
|
| Max | $200 / mo |
+1 more Show less
|
Is Mouse-1 worth the price?
Good value
The pricing is fair and offers good value for teams needing a managed AI agent harness.
This pricing is best for small to midsize development teams that want hosted cloud agents without the overhead of self-hosting.
Mouse-1 by the numbers
Toolradar data: Mouse-1's entry paid plan ($60/mo, billed monthly) ranks #213 of 303 (joint) for price among the AI Agents tools with a verified entry price and no stated unit, where the median is $29/mo. As of Oct 2026.
| Metric | Mouse-1 | Peer median | Position (1 is best) |
|---|---|---|---|
| Entry paid plan | $60/mo, billed monthly | $29/mo | #213 of 303 (joint) tools with no stated price unit, verified price |
| Free plan | No | 56% have one | n/a |
Each position is counted among the tools that have data for that row, as noted under it. A lower entry price ranks higher; prices are read on vendor pages and compared only with tools priced on the same basis. All AI Agents data · How we compute this. Advertising and placement fees are not inputs to any position.
Cite this: Toolradar, “Mouse-1 by the numbers”, Oct 2026.
Reviews
$99 Free with your review
Write a review Review Mouse-1, get a free AI guide
Share your experience and we will send you Improve Your Thinking Patterns Using ChatGPT, free.
Best Mouse-1 Alternatives
Top alternatives based on features, pricing, and user needs.
Portkey Freemium
Production stack for Gen AI builders: AI Gateway, Observability, Guardrails, Governance, and Prompt Management.
Patronus AI Paid
Reinforcement learning environments, tasks and training data for frontier AI labs.
LiteLLM Freemium
An AI Gateway for unified access, cost tracking, and fallbacks across 100+ Large Language Models.
Still deciding?
Most buyers shortlist 2 or 3 tools before committing. Pull a side-by-side comparison or browse the full alternatives shortlist below.
Explore More
Mouse-1 FAQ
How does Mouse-1 assist with content moderation in AI agent workflows?
Mouse-1 routes, gates, and verifies decisions in about 100 milliseconds, making it suitable for content moderation. It evaluates options and returns probabilities, allowing a frontier model to focus on reasoning while Mouse-1 handles rapid moderation decisions. This approach is up to 65 times cheaper than using frontier models for such tasks.
How does Mouse-1 differ from Portkey for agent decision tasks?
Unlike Portkey, which focuses on observability and routing for LLM calls, Mouse-1 is a dedicated decision model that plugs directly into an agent harness. It returns probabilities for each option in about 100 milliseconds, providing a specialized solution for routing, gating, and verifying agent decisions. This makes it up to 65 times cheaper than using frontier models for these rapid decisions.
What trade-offs should teams consider when adopting Mouse-1?
Mouse-1 is limited to decision-making tasks and cannot replace frontier models for reasoning or code generation. It requires integration into an existing agent harness, so it is not a standalone tool. Teams should expect to use it alongside a primary reasoning model to handle the hundreds of small, rapid decisions during agent tasks.
Which teams benefit most from integrating Mouse-1 into their agent workflow?
Teams building AI agents that require frequent, low-level decisions such as content moderation, document routing, browser state detection, or safety verification benefit most. These teams can use Mouse-1 to handle rapid decisions cheaply, allowing their frontier model to concentrate on complex reasoning and code generation.
How is Mouse-1 priced?
Mouse-1 uses a pay-per-use model based on input tokens with no charge for output tokens. A 2-day trial is available for the Pro and Max subscription plans, which also include hosted cloud agents, GitHub PRs, and higher usage limits. There is no permanently free tier.
How does Mouse-1 achieve high accuracy on unfamiliar tools?
Mouse-1 is designed to deliver high accuracy even on unfamiliar tools and repositories by processing the agent's current state and a set of options to return a probability for each. It operates with zero data retention, scoring each request and discarding it immediately, and it is resistant to prompt injection with a measured zero percent decision flip rate in tests.
Can Mouse-1 be integrated into any existing agent harness?
Mouse-1 plugs directly into an agent harness by taking the agent's current state and available options, then returning probabilities for each option. It requires integration into an existing agent harness and is not a standalone tool, but it is designed to work with any harness that needs rapid decision routing, gating, or verification.
Why is Mouse-1 suitable for safety verification in AI agents?
Mouse-1 is built to handle rapid decisions like safety verification in about 100 milliseconds. It returns a probability for each option, allowing the main frontier model to focus on reasoning. Its resistance to prompt injection, with a zero percent decision flip rate, makes it particularly reliable for safety-critical tasks.
How does Mouse-1 pricing compare with other AI Agents tools?
Mouse-1's entry paid plan ($60/mo, billed monthly) ranks #213 of 303 (joint) for price among the AI Agents tools with a verified entry price and no stated unit, where the median is $29/mo. Prices are read on each vendor's own pricing page and compared only with tools priced on the same basis. Source: Toolradar, Oct 2026.
Source: mouse.dev