Run this notebook: Open in Colab Open in Kaggle

Phase 15: AI Agents — Start Here¶

Build LLMs that can take actions — calling tools, browsing the web, writing code, and completing multi-step tasks autonomously.

What Is an AI Agent?¶

An agent is an LLM equipped with tools and a loop:

User goal
   ↓
[LLM] → decides action → [Tool call] → result
   ↑                                      ↓
   └──────────── observe + reason ─────────┘
   ↓
Final answer

Notebooks in This Phase¶

Notebook	Topic
`01_intro_to_agents.ipynb`	Agent concepts, tools, and the action loop
`02_function_calling.ipynb`	OpenAI and Anthropic tool/function use
`03_react_pattern.ipynb`	ReAct: Reason + Act + Observe
`04_agent_frameworks.ipynb`	LangChain agents, LlamaIndex agents
`05_multi_agent_systems.ipynb`	Multi-agent coordination and orchestration
`06_mcp_model_context_protocol.ipynb`	MCP — universal tool protocol for agents
`07_openai_agents_sdk_langgraph.ipynb`	OpenAI Agents SDK + LangGraph workflows
`08_reasoning_models.ipynb`	o1, o3, DeepSeek-R1 — extended thinking
`09_autonomous_agents_2026.ipynb`	Computer use, web agents, long-horizon tasks

Key Concepts¶

Concept	Description
Tool use	LLM calls functions (search, code, API)
ReAct	Interleaved Reasoning + Acting loop
Memory	Short-term (context) + long-term (vector DB)
Planning	Task decomposition and sub-task delegation
MCP	Model Context Protocol for tool standardization
Multi-agent	Specialized agents working in parallel

Prerequisites¶

LLM API basics (Phase 11: Prompt Engineering)
RAG Systems (Phase 08)
Embeddings (Phase 05)

Learning Path¶

01_intro_to_agents.ipynb         ← Start here
02_function_calling.ipynb
03_react_pattern.ipynb
04_agent_frameworks.ipynb
06_mcp_model_context_protocol.ipynb
07_openai_agents_sdk_langgraph.ipynb
05_multi_agent_systems.ipynb
08_reasoning_models.ipynb
09_autonomous_agents_2026.ipynb  ← Cutting edge