# Counterworld Documentation

> Technical documentation for building, running, controlling, and evaluating stateful software worlds for AI agents.

A Counterworld world is executable software with persistent state, business rules, interfaces, simulated users, history, and controllable time. It is not an LLM-generated imitation of an application. Agents act through the world's native interfaces, and evaluators inspect exact state changes and delayed outcomes.

## Start here

- [Documentation home](https://counterworld.dev/docs/): Documentation map and recommended reading order.
- [Quickstart](https://counterworld.dev/docs/quickstart.html): The shortest path from a world definition to an evaluated agent run.
- [Core concepts](https://counterworld.dev/docs/concepts.html): Definitions, state, actors, history, time, checkpoints, forks, and outcomes.

## World and runtime model

- [Worlds](https://counterworld.dev/docs/worlds.html): How definitions become populated, running software environments.
- [Runtime](https://counterworld.dev/docs/runtime.html): How agents interact while the runtime maintains shared state and valid mutations.
- [Simulation](https://counterworld.dev/docs/simulation.html): Simulated users, events, history, and controllable time.

## Evaluation and interfaces

- [Evaluation](https://counterworld.dev/docs/evaluation.html): Checkpoints, forks, state diffs, invariants, and delayed business outcomes.
- [TypeScript SDK](https://counterworld.dev/docs/sdk.html): Definitions, worlds, interfaces, time, checkpoints, forks, simulation, and evaluation.
- [Interface contracts](https://counterworld.dev/docs/reference.html): Agent, control, inspection, simulation, and evaluation surfaces.

## Optional

- [Product overview](https://counterworld.dev/): Non-technical overview of Counterworld.
- [World catalog](https://counterworld.dev/worlds.html): Ways to obtain or build a world.