Pokemon AI Agent with MCP Server

AI-powered Pokemon assistant using Claude AI and MCP protocol.

Architecture

User → Agent API → Claude AI → MCP Server → PokeAPI
            ↓
          Redis (conversation memory)

Run locally

Setup

python -m venv .venv
.venv\Scripts\activate        # Windows
# source .venv/bin/activate   # Mac/Linux

pip install -r mcp/requirements.txt
pip install -r agent-api/requirements.txt

Start Services

Terminal 1:

cd mcp
uvicorn src.main:app --port 8000

Terminal 2:

cd agent-api
uvicorn src.main:app --port 8001

Test

Open http://localhost:8001/docs

---

Docker Compose (uses Azure MCP)

Run

docker-compose up --build

Open http://localhost:8001/docs

---

API Usage

New conversation:

POST /pokemon_request
{"message": "Tell me about Pikachu"}

Continue conversation:

POST /pokemon_request
{"message": "What are its weaknesses?", "conversation_id": "from-previous-response"}

Available Tools

Tool	Description
get_pokemon	Pokemon info (types, stats, abilities)
get_pokemon_moves	List of moves a Pokemon can learn
get_move	Move details (power, accuracy, type)
get_type	Type effectiveness
get_ability	Ability description
list_pokemon	List Pokemon names