Skip to content
anythingai.space
AI Foundationsbeginner12 min · 9 steps

AI Agents and Their Design Patterns

From one augmented LLM call to workflows, autonomous agents and teams of agents, plus the guardrails that keep them safe.

CHATBOT · ONE TURNAGENT · LOOPS UNTIL THE GOAL IS METquestionactobservestateUserChat modelanswers onceGoalfix the failing testAgentplan · act · observeToolssearch · APIs · codeMemory
requestresponsetool calldataerrorsignal
1 / 9
ProblemStep 1 of 9

A chatbot answers once. An agent keeps going

Ask a chatbot to fix a failing test and it replies with a suggestion. The work is still yours.

An AI agent perceives its environment, decides, and takes actions toward a goal. It calls tools, reads the results, keeps state, and loops until the job is done, over many steps if needed.

Agents can follow predefined paths (workflows) or decide the path themselves (autonomous). This lesson walks from the simplest to the most autonomous.

InsightAutonomy, tool use, planning, state and feedback loops are what turn a model into an agent.
What changed
  • +Chat: one request, one reply
  • +Agent: goal → act → observe → repeat

Lesson 8 of 34 in AI Foundations