Agents
AI agents are systems where a model decides what actions to take, executes them with tools, and uses the results to determine what to do next. This section covers the fundamentals of agent design — from tool calling basics to multi-agent coordination and the guardrails that keep autonomous systems safe.
In This Section
What is an Agent
The agent loop (observe → plan → act → observe) and how it differs from a standard prompt-response interaction.
Tool Calling Patterns
How models invoke functions and APIs — defining tools, handling results, and designing reliable tool interfaces.
Multi-Agent Systems
Orchestrator and subagent patterns, handoff protocols, and when to split work across multiple specialized agents.
Guardrails & Permissions
Scoping what agents are allowed to do — permission models, confirmation gates, and rollback for destructive actions.
Agent Failure Modes
How agents go wrong — looping, tool misuse, context overflow, and prompt injection — and how to design against each.