What is a code agent?

• A. A type of agent that runs code to perform tasks and can integrate with libraries.
• B. A type of agent that uses only natural language prompts
• C. A type of agent that manages images
• D. A type of agent that schedules meetings

Answer: (A)

A code agent is defined by its ability to run real programming code to perform tasks, rather than relying solely on generating natural language. This means it can execute scripts, perform computations, manipulate data, and interact with external tools by importing libraries, calling APIs, and using the results of those computations to drive actions or produce outputs.

This capability makes it powerful for tasks that require actual processing, data analysis, or automation. By leveraging libraries, a code agent can handle complex operations, fetch and process data, or automate workflows in a repeatable way, rather than just describing steps or suggesting prompts.

The other ideas described—one that uses only natural language prompts, one that manages images, or one that schedules meetings—do not inherently involve executing code or integrating libraries to perform tasks. They might be capable in their own right, but they don’t capture the core ability that defines a code agent: running code to carry out tasks.