Ai Agent Bootcamp

AI Agent Bootcamp

A collection of progressively advanced AI agents built with the Anthropic Claude API. Part of my journey into AI engineering.

What's Inside

`hello_claude.py`

Basic API integration — single message request/response with token accounting.

`first_agent.py`

First multi-turn agent implementing the **ReAct pattern** (Reason + Act). Demonstrates:

• Tool definition with JSON Schema
• Agent loop with stop-reason handling
• Parallel tool execution
• Multi-turn reasoning

Tools: `get_weather`, `calculate`

`agent_v2.py`

Enhanced agent with production-grade features:

• System prompt for behavior/personality control
• Three tools including currency conversion
• Graceful error handling for unknown inputs
• Pre-emptive reasoning (agent skips tools when it knows they'll fail)
• Structured logging of each turn and tool call

Tech Stack

• Python 3.12
• Anthropic Python SDK
• python-dotenv for secret management

Setup

python3.12 -m venv venv
source venv/bin/activate
pip install anthropic python-dotenv

# Add your API key to .env
echo "ANTHROPIC_API_KEY=sk-ant-..." > .env

python agent_v2.py

Concepts Demonstrated

• ReAct (Reasoning + Acting) loop
• Tool use / function calling
• Parallel tool invocation
• System prompt engineering
• Error handling in agentic systems

What's Next

• Streaming responses
• Conversational memory
• Model Context Protocol (MCP) integration
• RAG (Retrieval-Augmented Generation)