Rotor Town

Typed decisions for real software

Rotor Town is a super-fast AI classifier.

Structured input. Typed answers. No generated prose.

Start buildingAPI docsSee how it works
A stitched felt town routes one input into three outcomes.

Measured, then questioned

Benchmarks

Hard accuracy across 2,000 typed decisions.

Read before comparing

Why this doesn’t matter

This benchmark reflects a specialist tuned for these workflows.

That is the point, not a universal victory.

Your domain should never depend on accidental training data.

Rotor Town defines it explicitly through versioned manifests.

Get better fit, lower costs, and faster decisions.

Skip traditional LLM fine-tuning.

Source and methodology

Not a traditional LLM

How it works

Discrete, typed inputs become decisions your code can use.

  1. 01

    Typed input

    Define fields, rules, and allowed values.

  2. 02

    Fast judgment

    We handle only the fuzzy decision.

  3. 03

    Workflow-ready result

    Receive choice, score, or noul with confidence.

Input
{
  "message": "Refund my order",
  "customer_tier": "pro"
}
Result
{
  "choice": "billing",
  "confidence": 0.96
}

Simple usage pricing

Pay for input.

Every account starts with $5.

$0.041per million input tokens256-token minimum per decision.JEV lists $0.042.Pricing details

Software is only the start

We’re here to help

We make AI work for your real workflows.

Build with us
Four Rotor Town team members as felt puppets.