LLM106 · Building AI Agents with Tools

Categories: AI & Automation
Wishlist Share

About Course

Build AI agents that use tools to look up data and take actions, with a controlled loop, clear permissions, human approval for risky steps and full logging.

Advance · 8 hours · 4 lessons, a lab, an assessed assignment and a portfolio piece

Who it is for

Developers comfortable with LLM APIs, structured outputs and RAG who want to automate multi-step business tasks safely.

Prerequisites

LLM104 and LLM105, or equivalent experience with structured outputs and retrieval.

What you will be able to do

  • Define tools with precise names, descriptions and JSON Schema parameters that a model can use reliably
  • Implement an agent loop with step limits, error handling and structured logs of every tool use
  • Apply least-privilege permissions and human approval for actions that change data or contact customers
  • Decide when a fixed workflow is better than an autonomous agent and justify the choice

Lessons

  • Lesson 1 · Tool use fundamentals
  • Lesson 2 · Agent loops, planning and memory
  • Lesson 3 · Safety, permissions and human oversight
  • Lesson 4 · Lab and assignment

Lab: An order-support agent with three tools

You will build an agent that answers order questions for a small online store by using three tools: order lookup in a SQLite database, shipping-status lookup from a CSV file, and a draft-reply tool that writes to a review queue instead of sending anything. You will add step limits, logging and an approval gate.

Free tools: Google Colab or VS Code, the Gemini API free tier or a tool-capable local model through Ollama, and local CSV and SQLite files

Assessed assignment

Design and build an agent for a small catering company that helps staff answer booking inquiries. It should check availability in a local calendar file, calculate a quote from a price table, and prepare a draft confirmation for a person to approve. Provide the tool definitions, the loop with limits, the permission model and logs from at least eight test runs, including two failure cases. Write a one-page design note explaining which steps you left to the agent, which you hard-coded as a fixed workflow, and why.

Criterion Weight What excellent work looks like
Tool design 25% Tools have clear names, descriptions and strict schemas; each does one job and returns structured, informative errors.
Agent loop and reliability 25% The loop has step and time limits, handles tool errors and recovers or stops cleanly.
Safety and permissions 30% Read and write tools are separated, risky actions require human approval, and the agent ignores instructions found in data.
Design reasoning and evidence 20% The design note clearly justifies agent versus workflow choices, supported by logs from realistic test runs.

Portfolio piece

A logged, permission-controlled business agent with documented tools, test runs and a design note.

Self-paced, 12-month access. Questions are answered in writing through the student panel. A completion record is issued when you finish the course.

Next step

Continue with LLM107 · Evaluating, Testing and Guarding LLM Applications · 8 hours · $56.00

Continue to the next course

Prefer the whole program? Program in AI Application Development & LLM Engineering · $378.00 (10% less than buying the courses separately)

Show More

What Will You Learn?

  • Define tools with precise names, descriptions and JSON Schema parameters that a model can use reliably
  • Implement an agent loop with step limits, error handling and structured logs of every tool use
  • Apply least-privilege permissions and human approval for actions that change data or contact customers
  • Decide when a fixed workflow is better than an autonomous agent and justify the choice

Course Content

LLM106 · Building AI Agents with Tools (8 h)
Build AI agents that use tools to look up data and take actions, with a controlled loop, clear permissions, human approval for risky steps and full logging.

  • Lesson 1 · Tool use fundamentals
  • Lesson 2 · Agent loops, planning and memory
  • Lesson 3 · Safety, permissions and human oversight
  • Lesson 4 · Lab and assignment
  • LLM106 Assessment · Building AI Agents with Tools
Scroll to Top