Learn Agent Architecture

Learn Agent Architecture explains that a model reasons but cannot act on its own; the harness around it runs tools, keeps state across calls, gates side effects and coordinates loops. It teaches the harness section by section, loop, tools, memory, permissions, context, tasks and interfaces, on the view that a coding tool, a chat assistant and an autonomous runner mostly differ in how they make these harness choices.

Each section follows the same four-part structure: the problem that layer solves, the general mechanism and control flow, how real systems implement it, and what breaks and how to mitigate it. Worked examples include Claude Code as a full frontier coding-agent harness, Hermes Agent for memory and always-on channels, and mini-swe-agent as the smallest complete agent loop, and readers are pointed to companion repositories that go deeper on agent memory and on one specific harness, deepseek-harness.