Agents SDK

Building agents from scratch teaches the mechanics. Using an SDK introduces structured abstractions for reasoning, tools, and memory. These components simplify implementation while preserving control over agent behavior.

• SDK mental model — Understanding the runtime concepts behind the SDK.
• Defining agents — Creating and configuring agents in code.
• Tools in the SDK — Registering and invoking callable capabilities.
• Memory in the SDK — Managing conversation state and persistence.
• Workflows in the SDK — Coordinating reasoning and actions across steps.
• Moving to the SDK — Migrating manual agents to SDK abstractions.