AI agent glossary
Agentic coding
Writing software by delegating multi-step tasks to a coding agent that reads files, edits code, and runs commands, while you direct and review.
In earlier AI-assisted coding, a model suggested the next few lines and you accepted or rejected them. In agentic coding, you describe a task and an agent carries it out across many steps: exploring the codebase, making edits in several files, running tests, reading the failures, and trying again.
Your role shifts from typing code to specifying work and judging results. Clear task descriptions, project instruction files, a plan you approve before edits begin, and a fast way to verify the outcome become the skills that matter most.
Agentic coding works best on tasks with a checkable definition of done, such as passing tests, a type check, or a visible change in the app. It still needs review: the agent can produce code that runs but is subtly wrong, insecure, or inconsistent with the rest of the codebase.
Instead of writing a new API endpoint yourself, you ask a coding agent to add it following the pattern of an existing one, write tests for it, and run the suite until it passes. You then review the diff before merging.
Learn more in the course
Related terms
Go from definitions to doing
Learn to use and build AI agents in 65 practical lessons.