Advanced AI Automation for Ragnarok Online

OpenKore AI — the Ragnarok Online bot powered by LLM decision-making, not just macros. A 3-tier AI brain (reflex → heuristic → LLM), multi-bot fleet coordination, AI agents, and an ML subconscious that learns from every outcome.

3-tier decision engine · 112+ API endpoints · 18 AI agents · Multi-bot fleet · Self-learning ML subconscious

Comprehensive Feature Categories

Organized into 10 specialized categories — updated with openkore-ai-v3

AI Decision Engine

Multi-Bot Fleet

Self-Learning ML

AI Agents

Cost Control

RO Mechanics

NPC Interaction

API Server

Live Console

Community

Featured Capabilities

What sets OpenKore AI apart — current as of openkore-ai-v3

3-Tier Decision Engine

Instant reflex rules (17 built-in) → heuristic scoring → LLM reasoning (DeepSeek, OpenAI, or local Ollama). Plans ahead via a PDCA loop.

Multi-Bot Fleet

Bots share state, auto-assign roles (tank/healer/dps/crafter), relay messages, and coordinate via a fleet coordinator.

ML Subconscious

Self-hosted behavior learning: shadow-mode training, macro distillation, and model promotion. Tracks outcomes per action/map/monster.

AI Agent Framework

18 specialized agents (combat, economy, navigation, questing, trading) managed by an orchestrator for whole-picture planning.

Cost-Controlled LLM

Per-bot daily token budget, hourly call limit, graduated tiers. Disable the LLM entirely and run reflex + heuristic.

112+ API Endpoints

FastAPI sidecar with 112+ endpoints across 19 routers: auth middleware, fleet coordination, telemetry, world data, PDCA autonomy loop.

Full RO Mechanics

Party, PVP, GVG, MVP, refine, cards, quests, crafting, job change, stat/skill allocation — Renewal-aware formulas.

Self-Adapting, Server-Agnostic

No hardcoded per-server solutions — per-server facts live in a DB-backed store, queried at runtime. Works with vanilla OpenKore configs.

OpenKore vs OpenKore AI

Vanilla OpenKore OpenKore AI
Macro-based decision makingReflex → Heuristic → LLM (3-tier)
Fixed behavior per configSelf-adapts from outcomes
Single bot per instanceMulti-bot swarm coordination
No cost controlPer-bot daily budget, graduated tiers
Community macros onlyCross-bot shared learning database
Reacts to conditionsPlans ahead via PDCA loop

OpenKore AI FAQ

What is OpenKore AI?

A modified OpenKore for Ragnarok Online with a 3-tier AI decision engine: reflex, heuristic, and LLM reasoning. It self-adapts from outcomes instead of following fixed macros.

Is it a Ragnarok Online bot?

Yes — an AI-powered bot that plans ahead via a PDCA loop, coordinates multi-bot fleets, and learns from its own outcomes.

Works with any server?

Server-agnostic. Works with any Ragnarok Online server that supports OpenKore, using standard control/config.txt format.

How is LLM cost controlled?

Per-bot daily token budgets, hourly call limits, and graduated tiers. You can disable the LLM entirely.

Multiple bots at once?

Yes — multi-bot fleet with shared state, auto-assigned roles, and fleet-coordinator coordination.

What is the 3-tier engine?

Reflex (instant safety rules) → heuristic (scored actions) → conscious LLM for cold-start plans. A trained ML subconscious drives skilled moment-to-moment combat.

Ready to let AI play for you?

OpenKore AI is open source — grab the code, point it at any RO server, and watch the 3-tier brain take over. Free to try with your own LLM.

GitHub Support the Development Create Account
Community research project. rAthena-AI world and openkore-ai-v3 are open-source proof-of-concept work — free to play, free to use, no item shop, no pay-to-win. Support is a voluntary tip that funds development time and project costs.