The Three Modes

Good news, everyone! We have three interfaces to talk about. Three ways to interact with Claude, each with a different philosophy, a different audience, and a different ceiling.

First: Chat. The original. The one you know. Single-threaded conversation, one prompt at a time, one response at a time. It is the foundation everything else is built on. You type, Claude responds, you type again. Simple, effective, and limited by the fact that you are doing one thing at a time in one context window.

Second: Co-Work. This is what we are here to learn. Visual agent orchestration with parallel execution, progress tracking, and a command center that lets you see what every agent is doing at any given moment. Co-Work takes the multi-agent patterns that power tools like Claude Code and makes them accessible through a visual interface. No terminal. No configuration files. No command-line flags. Just a UI that lets you spin up agents, assign tasks, and watch them execute.

Third: Claude Code. The terminal. Fully hackable. Maximum power. Claude Code is where everything is possible and nothing is hidden — you have direct access to the filesystem, the shell, git, MCP servers, custom hooks, and the full composability of a command-line tool.

Here is the part most people miss. Co-Work was literally vibe-coded using Claude Code in about a week. That is not a rumor. That is what happened. Claude Code is the source, and everything else — Co-Work included — is derivative. The reason that matters is not to diminish Co-Work. It is to establish the lineage. Co-Work takes the patterns that power Claude Code — parallel sub-agents, focused context windows, task decomposition — and wraps them in a visual interface that anyone can use.

That makes Co-Work a gateway. If you are not comfortable in a terminal, Co-Work lets you experience agent orchestration patterns — parallel execution, QA workflows, skill-based automation — without writing a single command. And once you have internalized those patterns, graduating to Claude Code is a natural next step, not a terrifying leap.

So what does the Co-Work command center actually give you? Let me walk through it because the interface is doing more than you might think at first glance.

Progress bars for every active agent. Not a spinner. Not "thinking..." Progress. You can see how far along each sub-agent is on its specific task, which matters enormously when you have four agents running in parallel — you need to know if one is stuck while the others are finishing.

Working folders for in-progress files. As agents produce deliverables — documents, code, analysis, presentations — those files appear in the working folder. You can open them, review them, and reference them without waiting for the final summary. This is a fundamental shift from Chat, where everything lives in the conversation thread and nothing persists as a file until you explicitly ask for it.

Context monitoring. You can see how much of the 200,000-token context window has been consumed. This matters more than you think, and we will cover it in depth in Lesson 8.

Pre-prompted buttons for common workflows. These are not just shortcuts — they encode best practices. "Research this topic," "Create a presentation," "Analyze this document" — each one structures the prompt in a way that is likely to produce better results than a freeform request.