AbadIA‑MCP: An AI Agent for The Abbey of Crime

AbadIA‑MCP bridges the gap between a modern AI assistant and a classic point‑and‑click adventure game, The Abbey of Crime (La Abadía del Crimen). By exposing the game as a Model Context Protocol server, it lets developers treat the entire game world as a set of high‑level tools that an AI can call. The server hides the low‑level input mechanics (key presses, mouse clicks) behind REST endpoints that encapsulate complex actions such as “move to location” or “investigate object.” This abstraction allows an AI agent—built with Google’s Agents Developer Kit—to concentrate on strategic decision‑making, narrative progression, and puzzle solving without wrestling with the game’s internal state.

The core problem addressed by AbadIA‑MCP is seamless AI integration with legacy game engines. Classic games typically expose only rudimentary input hooks, making it difficult for modern language models to interact meaningfully. AbadIA‑MCP solves this by providing a comprehensive, real‑time game state API () and a suite of high‑level commands that map directly to in‑game actions. Developers can therefore craft agents that reason about the plot, plan multi‑step investigations, and navigate the environment using declarative instructions rather than raw keystrokes.

Key capabilities include:

• High‑level command execution: Endpoints such as and encapsulate pathfinding, animation sequencing, and interaction logic.
• Real‑time state monitoring: The server continuously exposes the full game state, enabling agents to react to dynamic events (e.g., new clues appearing).
• Modular architecture: The server and agent are decoupled; new skills or game modules can be added without altering the core logic.
• ADK integration: The agent leverages Google’s ADK for planning, tool use, and prompt management, ensuring that the AI remains focused on high‑level goals while delegating execution to the server.

Typical use cases span both research and entertainment. In academic settings, researchers can study how language models solve spatial puzzles or follow narrative arcs in a controlled environment. Game designers might prototype AI‑driven companions that guide players through complex puzzles or test procedural generation of game content. Hobbyists can experiment with “AI‑controlled” gameplay, creating novel experiences such as cooperative puzzle solving or automated walkthroughs.

Integration into existing AI workflows is straightforward: an MCP‑compatible client (e.g., Claude, OpenAI’s tools API) sends tool calls to the AbadIA‑MCP endpoints. The agent’s prompt includes persona and goal definitions sourced from , ensuring that the AI's decisions align with the game’s narrative. Because the server handles all low‑level interactions, developers can focus on refining strategy, enhancing natural language understanding, or extending the game’s logic without touching the core engine.

In summary, AbadIA‑MCP transforms The Abbey of Crime into a rich, programmable environment for AI agents. By abstracting complex gameplay mechanics behind a clean MCP interface and pairing it with a strategy‑oriented agent, it offers developers a powerful platform for exploring AI behavior in interactive storytelling and puzzle‑solving contexts.