WorldFork Documentation

WorldFork is agent-operated social simulation infrastructure for branching timelines. It is best understood as a Monte Carlo tree search of the real world: initialize a scenario, explore plausible branches, score terminal outcomes, and compare the resulting timelines. It exposes a FastAPI backend, Celery workers, durable Postgres and Redis state, an artifact store, and the worldfork CLI for agent and operator workflows.

Wiki And Documentation Mirrors

Resource

Use this when

Read the Docs

You want the maintained documentation site generated from this repo.

Deep Wiki

You want a DeepWiki view of the repository and codebase.

The documentation is organized around the jobs people actually need to do:

Start here

Use this when

Setup

You need a local backend, CLI, and agent skill installed

CLI

You need command examples or agent-safe operating patterns

Demos

You want to run Atlas or the live runtime smoke

Architecture

You need the runtime model and storage layers

Runtime Walkthrough

You need the exact initializer, tick, queue, and observability flow

Reporting

You need report versioning, artifacts, and continuations

Agent Interface

You are writing or operating an AI agent against WorldFork

Guides