AI that talks, plans, and acts. AgentCore turns natural language into real Unity gameplay with full memory, planning, and control — all demonstrated in 3 interactive demo scenes.Talk. Plan. Play. — Now your game understands you.With AgentCore, your Unity world becomes truly interactive — powered by AI that listens, plans, and acts.AgentCore is an AI Agent Framework that bridges natural language and real gameplay.Say "Go to the shop and buy three potions," and watch your AI-controlled character move, talk to the NPC, and make the purchase — automatically.No hardcoded triggers. No dialog trees. Just intelligent behavior driven by language.What Makes AgentCore DifferentUnlike simple "ChatGPT-in-Unity" plugins, AgentCore provides a full cognitive pipeline:Intent Recognition → Multi-Step Planning → Unity Command Execution → Memory Recall.It doesn't just reply — it remembers, understands, and acts.Whether you're building an interactive RPG, an AI companion, or an experimental simulation, AgentCore lets your AI agents think and behave like real players.Key Features1. Context-Aware Memory SystemAgentCore remembers conversations, past actions, and player context.Even after long play sessions, it recalls history naturally:"What did I buy last time?"→ "You bought three potions and a sword. Want to visit the blacksmith again?"Its built-in summarization keeps memory coherent and efficient.2. Direct Unity Command ControlThrough the CommandRegistry, AI directly triggers Unity functions such as:MoveCommand → Character movementInteractCommand → NPC or object interactionBuyItemCommand → Store transactionsEvaluateArgumentCommand → Logical evaluationYou describe the action — AgentCore executes it inside Unity.3. Multi-Step AI PlanningAgentCore's dependency-based planner turns complex goals into actionable sequences:"Go to town, talk to the merchant, and buy a sword."→ Step 1: Move → Step 2: Interact → Step 3: PurchaseSupports sequential and parallel task execution — ideal for RPGs, sims, or AI strategy prototypes.4. Dual Operation ModesUnity Standalone Mode: Works instantly, no setup needed — great for prototypes and indie projects.Python Server Mode (LangChain): Full autonomy and custom agent logic using FastAPI & LangChain.Start lightweight, scale to advanced AI systems anytime.5. Built for Learning & ResearchAgentCore is also an educational playground for AI developers.Learn real-world agent design inside Unity:Intent → Plan → Execute → Memory pipelineLangChain-based reasoning & memoryUnity ↔ Python communication via FastAPIModular scene-based command designPerfect for students, AI engineers, and game devs exploring agentic gameplay.3 Fully Playable Demo Scenes1) RPG Village — "AI Understands the World"Say "Go to the potion shop and buy three potions."The AI navigates, interacts, and trades — a complete autonomous gameplay experience.2) Visual Novel — "AI Understands Emotion""Give Sakura a gift and ask how she feels."AI reads context, tracks emotional state, and responds dynamically like a real character.3) Debate Battle — "AI Judges Logic""Let the AI referee decide who wins."AI evaluates arguments, generates rebuttals, and scores reasoning in real time.Who It's ForIndie Game Developers: Bring NPCs to life with true conversational intelligence.AI Researchers & Educators: Test planning, memory, and reasoning inside a real 3D environment.Prototype Builders: Rapidly build AI-driven gameplay concepts without custom backends.LangChain Users: Connect Python-based AI agents directly to Unity.AgentCore turns Unity into a living AI laboratory where imagination becomes behavior.Why Developers Love AgentCoreNatural language → Real game actionAI planning, reasoning, and memory all in one frameworkDual modes: Quick-start (Unity) or advanced (Python LangChain)Works with multiple AI providers and local modelsThree complete, ready-to-run demo scenes includedSupported Unity Version: 6000.0.58f2 or higher (Unity 6 LTS recommended)Languages: C# / Python (optional)AI Providers: OpenAI (GPT-4/3.5), Claude 3.5, Gemini 1.5, Local LLMs (Ollama / LM Studio)OS Compatibility: Windows / macOS / LinuxMemory System: Context summarization + persistent conversation historyPlanner: Dependency-based multi-step executionNetworking: UnityWebRequest ↔ FastAPI (Python server mode)Dependencies: TextMeshPro, Newtonsoft.JsonPython Requirements: FastAPI, LangChain, OpenAI / Claude / Gemini SDKsTested Models: GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro, Llama-3-8B (local)This package was partially created with the assistance of AI tools.Specifically, ChatGPT was used to generate sample character images for the demo scene and to support idea development for architecture and code structure. Claude was used to help review and clean up portions of the source code for clarity and consistency.All final content — including code, logic, and assets — was manually written, reviewed, and tested by the developer to ensure originality, quality, and full functionality.
