Skip to content
TextView.AI
Sign inGet started
/
14 chapters · about 37 min · Beginner to intermediate

Introduction to agentic systems

TextView.AI
Start course

Learn what AI agents are, when to build one, and how to build a crew of agents in Python that reads a contract, flags the risks, and checks its own work.

About#

Build production-grade multi-agent systems in Python across fourteen practical chapters. Learn when to deploy autonomous agents over single prompts or deterministic workflows, how to structure specialist teams, and how to combine language models with unit-tested code.

Throughout this course, you build a three-agent crew that analyzes commercial vendor agreements, extracts key terms, flags financial risks, and audits its own citations against source documents.

What you'll learn#

  1. Agent architectures: Differentiate autonomous agent loops from single prompts and fixed pipelines like RAG.
  2. Core building blocks: Implement role, focus, tools, cooperation, guardrails, and memory.
  3. Multi-agent orchestration: Structure and execute collaborative multi-agent teams using CrewAI.
  4. Declarative configuration: Externalize prompt text, agent personas, and task definitions into YAML.
  5. Typed results: Enforce runtime validation and structured Python objects with Pydantic schemas.
  6. Custom tools: Build, unit-test, and deploy deterministic Python tools for calendar arithmetic.

Prerequisites#

  • Python: Familiarity with Python classes, functions, and type hints.
  • Environment: Experience creating virtual environments and running terminal commands.
  • LLM Access: An OpenAI API key or a local Ollama instance.

No background in machine learning is required.

Technology stack#

  • Python: 3.13 (supports 3.10 through 3.13)
  • CrewAI: crewai 1.15.28 and crewai-tools 1.15.28
  • Pydantic: 2.12.5
  • PyYAML: 6.0.3
  • pytest: 9.1.1

All code samples are verified against these exact package versions.

Frequently asked questions#

What does it cost to follow along?#

With Ollama, execution is completely free. With hosted models like OpenAI gpt-4.1-mini, running the finished crew costs less than a few cents per run (~14,000 tokens).

Why CrewAI?#

CrewAI's primitives (Agent, Task, Crew, Process) map directly to core multi-agent architectural principles. The concepts you learn transfer seamlessly to LangGraph, AutoGen, and custom Python agent loops.

Can I review custom agreements?#

Yes. Save any agreement as Markdown or plain text in the contracts/ directory and pass its path to kickoff(inputs={"contract_path": ...}).