Backyard

A comprehensive guide to building an AI agent with tools, skills, and memory - everything you need before going to production

Learning Claude Code's agent spawning by inspecting its leaked source code.

Learning Claude Code's tool system by inspecting its leaked source code.

Learning Claude Code's App State by inspecting its leaked source code.

Learning Claude Code's context compaction by inspecting its leaked source code.

Learning Claude Code's permission system by inspecting its leaked source code.

Learning Claude Code's multi-agent coordination by inspecting its leaked source code.

Anthropic invented MCP a year ago. Now they're experimenting with skill-ifying it. Here's what the leaked source reveals about the pivot.

Learning Claude Code's query engine by inspecting its leaked source code.

Learning Claude Code's memory system by inspecting its leaked source code.

Event-driven architecture, multi-agent orchestration, scheduling, and memory systems - everything you need to run agents at scale.

To vibe or not to vibe, that's the question. Draw the line between throwaway AI vibe coded prototypes and sustainable software engineering without losing your edge.

Stop memorizing new terminology every time a new tool launches. Understand how agents manage context, use prompts, and call tools, and everything else starts making sense.