Title here
Summary here
Module · 5 lessons
Planning and Reasoning
Help your agent tackle hard, multi-step tasks — decompose goals into steps, interleave reasoning with action (ReAct), and let the agent reflect on and fix its own work.
At a glance
Level
Intermediate
Lessons
5 lessons
Time to complete
1 week
Cost
Free forever · no sign-up
Welcome to Planning and Reasoning, the fifth module. Your agent can already loop, call tools, validate inputs, and remember — but on genuinely hard tasks, a bare loop can still wander: calling tools in a muddled order, missing a step, or confidently producing a flawed answer. The fix isn’t a bigger model; it’s giving the agent better reasoning structure. This module covers the three patterns that do exactly that.
You’ll start by seeing why agents need to plan — and why making reasoning explicit helps on multi-step work. You’ll learn task decomposition: breaking a big goal into ordered sub-tasks the agent can actually carry out. You’ll meet ReAct, the pattern of interleaving reasoning with acting so each tool call is a deliberate choice, not a guess. And you’ll add reflection: a “check your own work” step where the agent critiques a draft and revises it before finishing. The module ends by giving Atlas all three — a planning, reasoning, self-correcting trip planner.
Every pattern’s orchestration here — plan-then-execute, the ReAct trace, the reflect-and-revise loop — is real, runnable Python, verified end to end against the agent loop you’ve built. Start with Lesson 1 on why agents need to plan.
Lessons in this module
1
Why Agents Need to Plan
See why a bare tool loop wanders on hard tasks — and how making the agent's reasoning explicit, interleaved with its actions, turns guesses into deliberate steps.
2
Task Decomposition
Break a hard goal into an ordered list of steps, then execute them one at a time — the plan-then-execute pattern that turns a vague request into a sequence the agent can actually carry out.
3
ReAct: Reasoning and Acting
Interleave reasoning with action: make the agent think before each tool call, observe the result, and think again — the ReAct loop, built on the agent loop you already have.
4
Reflection and Self-Correction
Add a 'check your own work' step: the agent critiques its own draft against the task, then revises — catching missed constraints before it finishes, in a reflect-and-revise loop.
5
Guided Project: A Planning, Reasoning Atlas
Combine all three patterns on one agent: Atlas decomposes a hard trip request into steps, reasons and acts (ReAct) through the steps that need facts, then reflects on its draft and revises before finishing.
Achievement
Complete all 5 lessons to finish the Planning and Reasoning module.