What is Claude Code?
The Harvard Chan AI Glossary
In our classes, we often pause to explain the ideas behind AI and the language used to describe it. These posts grew out of those conversations. They’re for anyone who would like to understand AI better, with no technical background assumed.
Claude Code is Anthropic’s agentic coding tool. It can work with project files, propose and make code changes, run commands and connect to development tools in supported environments. Code is the set of instructions a computer executes. The product is designed to work on those instructions within a project, rather than merely discuss them in an ordinary conversation.
An AI model is the trained component that interprets the assignment and proposes responses. Claude names a model family as well as an assistant application. Claude Code is the coding-agent product around model capabilities. An agent is a system that can choose actions, observe results and continue toward a goal.
Suppose a website’s search function fails on names containing apostrophes. A coding agent could inspect the relevant files, propose a repair and run tests. Tests are checks of expected behavior. A useful test would include the failing name and ordinary searches that already worked, so that fixing one case does not quietly break another.
The surrounding harness is the software coordinating the model, tools and task progress. Tools are callable functions such as reading a file or executing a command. Permissions determine which operations are available and when approval is required. A request to investigate a bug should not automatically imply permission to publish a new version.
A command-line interface is a text-based way to operate software. An integrated development environment is a workspace combining editing, file navigation and execution. Claude Code’s supported interfaces can place the agent in different working settings; the interface is not a separate model.
Treat Claude Code as a harness for doing work. Pay attention to context—the information actually supplied to the agent—and to the evidence that its work succeeded.
For a clinic follow-up prototype, an assignment might be to fix a bug that shows a canceled appointment as still upcoming. Using fictional records, the team could ask Claude Code to investigate, make a bounded repair and run checks for both canceled and active appointments. The useful evidence is the changed behavior and test results, not the agent saying “fixed.” Editing the code, approving the change and releasing it into a live service remain separate responsibilities.
Source: Anthropic — Claude Code Overview.
If you would like to find out more about AI and improve your leadership skills, consider the AI in Health Care Certificate of Specialization from the Advanced Learning Academy at Harvard T.H. Chan School of Public Health. You can take an individual course or work toward the full certificate.
AI is a fast-moving field. This explanation reflects our class discussions and related material at the time of writing. Terminology, capabilities and tools may change.
For suggested additions or corrections, please email me at tpanch@hsph.harvard.edu.
Dr. Trishan Panch
Program Director, AI in Health Care
Harvard T.H. Chan School of Public Health — Advanced Learning Academy