v2.3.0

AI RPG Engine Build worlds. Simulate them. Improve them.

The first RPG engine designed for experimentation. Deterministic simulation runtime + AI-assisted design studio + six-pillar tactical combat.

Get started Read the Handbook

Install

npm install -g @ai-rpg-engine/cli

Studio

ai chat

Onboard

/onboard

Core Capabilities

What makes AI RPG Engine different.

Deterministic Simulation

Tick-based engine with world state, events, perception, cognition, faction beliefs, rumor propagation, and seeded RNG. Every run can be replayed exactly.

AI-Assisted Worldbuilding

Scaffold rooms, factions, quests, and districts from a theme. Critique designs, repair schema errors, and guide multi-step builds. AI suggests — you decide.

Replay Analysis

Structured findings explain why events happened, where mechanics break down, and which systems create instability. Analysis feeds directly into tuning.

Experiment-Driven Balancing

Run batches of simulations across seeds. Detect variance, sweep parameters, compare tuned vs baseline. Turn world design into a testable process.

Studio Workflow

CLI design studio with dashboards, issue tracking, experiment browsing, session history, guided onboarding, and context-aware command discovery.

Six-Pillar Combat

Guard, brace, engagement, interception, precision-vs-force dimensions, and AI tactics. Stats shape every formula. buildCombatStack wires it in 7 lines.

27 Built-In Modules

Combat, tactics, engagement, abilities, dialogue, cognition, perception, factions, rumors, districts, progression, environment, and more. All composable, all deterministic.

Genre-Agnostic

Same core runs dark fantasy, cyberpunk, detective noir, or any setting. 10 starter worlds show composition patterns to learn from and remix. Genre belongs to content packs, not the engine.

Quick Start

Install

npm install -g @ai-rpg-engine/cli

Start the studio

ai chat
/onboard

Build a world

create-location-pack haunted chapel district
critique-content
simulate

Analyze and improve

analyze-balance
tune paranoia
experiment run --runs 50

Built-In Modules

27 composable simulation modules.

Module

Description

combat-core

Hit/damage formulas, guard, defeat, stat-mapped combat

combat-tactics

Brace and reposition actions with AI scoring

combat-states

Guarded, off-balance, exposed, fleeing — with narration

combat-intent

AI action selection with 8 tactical intents and pack biases

combat-resources

Stamina costs, resource gains/drains, AI resource awareness

combat-roles

Boss phases, role templates, encounter builders

engagement-core

Engaged/protected/backline/isolated zones, frontline collapse

defeat-flow

Morale cascades, flee/surrender thresholds, defeat narration

combat-recovery

Post-combat HP/morale/resource restoration

combat-review

Traced formula explanations for balance analysis

dialogue-core

Graph-based dialogue trees with conditions

ability-core

Abilities with cooldowns, costs, targeting, and AI intent

ability-effects

Status application, healing, damage, stat modification

inventory-core

Items, equipment, use/equip/unequip

traversal-core

Zone movement and exit validation

status-core

Status effects with duration and stacking

environment-core

Dynamic zone properties, hazards, decay

cognition-core

AI beliefs, intent, morale, memory

perception-filter

Sensory channels, clarity, cross-zone hearing

narrative-authority

Truth vs presentation, concealment, distortion

progression-core

Currency-based advancement, skill trees

faction-cognition

Faction beliefs, trust, inter-faction knowledge

rumor-propagation

Information spread with confidence decay

district-core

Spatial memory, zone metrics, alert thresholds

belief-provenance

Trace reconstruction across perception/cognition/rumor

observer-presentation

Per-observer event filtering, divergence tracking

simulation-inspector

Runtime inspection, health checks, diagnostics

Design Workflow

Scaffold content

create-location-pack --theme "abandoned mine" \
  --factions miners_guild,deep_crawlers

Analyze balance

simulate
analyze-balance
suggest-fixes

Tune mechanics

tune rumor propagation
tune-step
tune-status

Run experiments

experiment run --runs 50
experiment compare baseline tuned
/findings