Claude Code is the agent we use in this course. Despite the name, it is not only for programmers. It is a general-purpose agent that happens to be very good with files, and your work is made of files.
Take it apart with the four-part picture:
- Model: one of Anthropic’s Claude models. You can switch between them.
- Harness: the Claude Code program itself. It runs the loop, shows you what the model wants to do and asks before it changes anything, according to the permission mode you set.
- Tools: reading and writing files, running programs, searching the web, and connectors to other systems you add yourself.
You can use it from several windows, all onto the same agent: a terminal, the Claude desktop app, an extension in VS Code, a plugin in JetBrains, or the web. Pick the one you find comfortable. The course shows the terminal and VS Code side by side.
Example
Open a folder with last month’s meeting notes, start Claude Code there and type: “Read all the notes and list every decision that still has no owner.” It reads the files, thinks, and answers with a list, citing which note each item came from.
Why it matters for you
Anthropic sells the same model in three products, see Chat, Work (Cowork) and Code. Chat is an advisor on the phone. Work is a contractor you hand a folder and a task. Code is a colleague at the next desk with access to what you have. The difference is the tools: only Code can run the programs on your own computer, with your logins. Chat and Work run programs only in a sandbox. Once you are comfortable at the Code level, every other AI product is a simpler version of the same thing.
Related
- Agent, Harness, Tool - the parts it is made of
- Project and Instructions file - the folder it works in and the rules it reads there
- Choosing a client and Your first session - installing it and the first twenty minutes
- Claude Code cheat sheet - keys, commands and modes on one page
- Roadmap - when we install it and what we do first