Most people treat Claude Co-work like a smarter chat window: ask a question, get an answer, move on. In this video, I show why that mindset leaves a ton of value on the table. Co-work is a desktop agent—it can access your files, connect to your tool stack, run multi-step work while you’re away, and even spin up parallel sub-agents to knock out big tasks faster. I walk through 25+ practical tips that we actually use, not theory. I cover when to use Chat vs Co-work (tokens matter), how to turn on memory, and how to set global instructions so every session behaves the way you want. Then we get into the workflow stuff that saves hours: managing plugins/skills/connectors from the “Customize” briefcase, working across multiple folders, running Co-work with no folder selected, and using a folder-level claude.md for project-specific rules. I also show slash commands, Dispatch for assigning tasks from your phone, reviewing Claude’s execution plan before destructive actions, duplicating sensitive files, and keeping your computer awake so scheduled/long tasks don’t die. On the automation side, we talk connectors (Gmail/Notion/Calendar), installing and editing plugins, finding additional plugins online, using parallel sub-agents, scheduling tasks two ways, using the Ideas library for pre-built prompts, and batching related work into one session to avoid reloading context. If you want Co-work to feel like an ops teammate instead of a chatbot, these tips are the playbook.