Track
ai-agents
AI Agents
36 lessons: tool loops, memory, guardrails, evals, and production agents—read-focused practice with local Python checks and 108 MCQs.
- Mode
- none
- Practice
- Read / quiz
- Lessons
- 36 units
Before you start
AI agents: a model in a loop that observes, chooses a tool, reads the result, and stops on a rule you wrote down.
A chat reply is not an agent. Teams that ship tool use without permissions, traces, and an off switch ship incidents.
Support workflows, research assistants, code agents, and any product that lets a model call an API.
Read each lesson, write the allowed actions and the failure case, run the short Python checks locally, and answer the MCQs.
After /ai/intro and /gen-ai/intro—when you are about to let a model take actions, not only draft text.
Lesson order
Sequential — follow top to bottom36 lessons are live in this track. Start from step 01 for the smoothest path.
-
01 intro What an AI agent is
beginner
Open → -
02 agent-loop The observe-plan-act loop
beginner
Open → -
03 tools-and-actions Tools are the action space
beginner
Open → -
04 memory Short memory and long memory
beginner
Open → -
05 planning Breaking a task into steps
beginner
Open → -
06 react-pattern ReAct: reason, then act
beginner
Open → -
07 function-calling Structured tool calls
intermediate
Open → -
08 multi-step Tasks that take many steps
intermediate
Open → -
09 state-machines Graphs when the path is known
intermediate
Open → -
10 human-approval Human approval gates
intermediate
Open → -
11 guardrails Boundaries around actions
intermediate
Open → -
12 evaluation Scoring an agent
intermediate
Open → -
13 traces Traces and observability
intermediate
Open → -
14 errors Tool errors and retries
intermediate
Open → -
15 cost Token and tool cost
intermediate
Open → -
16 retrieval-agents Agents that search knowledge
intermediate
Open → -
17 computer-use Computer and browser actions
advanced
Open → -
18 code-agents Agents that edit code
advanced
Open → -
19 research-agents Research workflows
advanced
Open → -
20 multi-agent When several agents help
advanced
Open → -
21 handoff Handing off to a specialist
advanced
Open → -
22 protocols How agents talk to tools
advanced
Open → -
23 schemas Schemas for tool arguments
advanced
Open → -
24 idempotency Safe retries
advanced
Open → -
25 permissions Least privilege
advanced
Open → -
26 sandbox Isolating side effects
advanced
Open → -
27 testing Deterministic tests
advanced
Open → -
28 datasets Task suites for evals
advanced
Open → -
29 production Shipping an agent
advanced
Open → -
30 monitoring Live failure modes
advanced
Open → -
31 security Prompt injection through tools
advanced
Open → -
32 privacy What the agent is allowed to see
advanced
Open → -
33 ux Showing work to the person
advanced
Open → -
34 failure-recovery Getting unstuck
advanced
Open → -
35 interview Interview review
advanced
Open → -
36 capstone Design a complete agent
advanced
Open →