Projects Pokémon TCG Agentic AI

Pokémon TCG Agentic AI

An autonomous agent that plays the Pokémon Trading Card Game — reasoning through board state, hand options, and turn strategy to make competitive plays.

Competition Agentic AI

Tech Stack

  • Python
  • LangChain
  • Agents
  • LLM
View Competition
TCG Agent Board dashboard showing game status, board state, hand options, AI reasoning trace, and action candidates for Pokémon TCG agentic play

Overview

Built for the official "PTCG AI Battle Challenge Strategy" competition hosted by The Pokémon Company on Kaggle. The challenge: design an agent capable of reasoning through Pokémon TCG gameplay — evaluating board state, hand composition, energy management, and matchup dynamics to make strategic, rules-legal decisions turn by turn.

What I Built

  • Designed an LLM-based agentic decision loop that parses game state and selects legal, strategically sound actions each turn
  • Implemented reasoning over core TCG mechanics: type matchups, energy attachment, retreat cost, and prize card trading
  • Built evaluation logic to compare agent performance across different game states and opponent strategies
  • Entered into an official Pokémon Company-sponsored Kaggle competition, testing agentic reasoning against a real competitive ruleset