The Agent Loop
The Loop: Perceive → Act
The Agent is not magic. It is an infinite loop.
1. Perceive
User: "What's the weather?"
2. Think (LLM)
Context: System Prompt + Tools + User Message Thought: "I need to check the weather tool."
3. Act
Call: web_search("weather nyc")
4. Observe
Result: "75°F, Sunny"
5. Reply
Agent: "It is currently 75°F and sunny."
This loop repeats for every interaction.
Knowledge Check
What comes after the Agent 'Acts' (calls a tool)?