Claude Code Loop Guide: /loop, /goal, and the Cron Loop Behind 1,800+ Unattended Runs (2026)
Claude Code loop guide: how /loop and /goal work, the jitter and 7-day expiry most guides skip, and the cron loop behind 1,800+ unattended runs.
Claude Code loop guide: how /loop and /goal work, the jitter and 7-day expiry most guides skip, and the cron loop behind 1,800+ unattended runs.
I checked my AI agent instructions against the live container. 7 claims had drifted and 4 blanks were never filled. Here’s the 10-minute Sunday check.
MCP vs API isn’t a rivalry. MCP sits on top of an API. Here’s when the extra layer pays off, what it costs, and why my agents running 10+ businesses make zero MCP calls.
Before a vendor tells you to rotate every key, map where each one lives. My container held 33 secret names, 29 real secrets, and 4 keys stored twice.
An AI coding agent plans, edits and runs code on its own. Here’s the four-level autonomy ladder, the headless test every ‘best of’ list skips, and 30 days of real unattended runs.
Is a proactive AI agent worth it? 150 scheduled check-ins from my fleet, what they found, what they may do without asking, and where findings should land.
MCP Apps put interactive UI inside Claude and ChatGPT. What they are, where they run, and the operator test for when a workflow needs a screen at all.
Every evening one of my agents picks five blogs, matches each one to a post of mine they might want to link, and loads them into a cold email campaign. It runs on the Instantly…
Do you need an MCP gateway? What it does, the July 2026 spec change, real prices, and the threshold from a fleet of 10+ agents running without one.
Every Wednesday I share one lesson from running my own fleet of AI agents. This week’s is about the safety net everyone tells you to add. A human in the loop AI agent setup means…