Chapter C64 · Interactive Course
Meet Claude Code and start a session
Model against agent program, surfaces and execution location, setup stages, the right checkout and the first prompt.
Core Learning Concept
Plan a first Claude Code session using the fictional project and setup information. Identify the surface, execution location, account requirements, and intended workspace. Begin with an explanation grounded in files, then inspect the first proposed action. This course practice does not require installing software or entering real credentials.
What You’ll Learn in This Chapter
-
Claude Code is an agent product
Separate Claude model capability from the software using tools.
Claude Code is an agent product around a model, with capabilities for reading code, editing files, and using development tools under its configured permissions. The product and the model are distinct. A successful task depends on the workspace, available tools, instructions, and verification as well as model capability.
-
Choose the surface and execution location
Identify where a selected session accesses files and runs work.
A surface is the interface used to access the agent, such as a terminal, editor integration, desktop app, or browser. Execution location tells you where the work runs. These are separate questions. Choose a supported setup that gives the task its required files and lets you inspect the resulting changes.
-
Setup includes account and environment
Recognize installation, sign-in and project selection as separate steps.
Setup includes a supported installation or surface, an appropriate account or provider configuration, and the project’s own dependencies. Requirements vary by environment and product version. Use current official setup guidance and keep credentials in the supported secure flow. A successful login alone does not prove the project can run.
-
Open the intended project
Verify the working folder before a request.
Open the target project and inspect its starting state before requesting edits. Confirm the folder path, repository, active branch, and existing changes to avoid modifying the wrong directory and losing work.
-
Begin with a read-only orientation
Ask for a source-backed explanation of a small feature.
A read-only orientation asks the agent to explain a specific code path using files and references before modifying it. For example, trace how the form submits its data. Check that the explanation matches the source. This establishes useful context and can expose missing dependencies or assumptions early.
-
Inspect the first proposed action
Connect a permission prompt to the concrete command.
Inspect what the proposed action will touch and how it serves the task. Reading a file, editing it, and running a command have different effects. Use the product’s permission controls and the task’s existing authorization. When a consequential decision is needed, review the concrete operation rather than only a broad intention.
-
First session storyboard
Sequence setup, workspace verification and a source-backed read-only explanation from supplied screens, and say what each one establishes.