Redline lets Claude Code post its review of a diff as comment threads on the lines in VS Code. You answer on the line, submit, and Claude fixes the code and replies in place.

Headless Claude with --json-schema, --agents, and --allowedTools becomes a callable AI subroutine. I used it to build Memory Diagnoser, a tool that audits the instruction files feeding my own agents.

Plugin monitors are an experimental Claude Code feature: a long-lived process whose stdout becomes task notifications. How I used them in two projects, a kanban board and a message bus.

A browsable teardown of Claude Code's two security tools, and the six design lessons underneath them.

A landscape map of browser automation tools for AI coding agents in mid-2026 — Playwright CLI, Vercel's agent-browser, Stagehand, Browserbase, ReportPortal, and where Claude for Chrome fits.

A competitive-landscape report of four GitHub repos across agent context engineering, agent memory, and data cataloging — where they overlap, where the gaps are.

A durable message bus that lets independent Claude Code instances in different tmux windows coordinate and hand work to each other.