The agentic loop

When you give Claude a task, it works through three phases: gather context, take action, and verify results. These phases blend together — Claude uses tools throughout, chaining dozens of actions and course-correcting along the way.

The agentic loop, redrawn from the official docs in Patterson brand colors.

The loop adapts to the ask. A question about your codebase might only need context gathering; a bug fix cycles through all three phases repeatedly. You are part of the loop too — press Esc to interrupt and steer at any point.

The loop is powered by two components: models that reason and tools that act. Claude Code is the agentic harness around the model: it provides tools, context management, and the execution environment that turn a language model into a capable coding agent.

Watch it happen

Ask Claude to "fix the failing tests" and watch the loop: run tests → read errors → search sources → read files → edit → re-run tests. Each tool result feeds the next decision.

Check your understanding

Question 1What are the three phases of the agentic loop?

APlan, code, deployBGather context, take action, verify resultsCRead, write, commitDPrompt, respond, repeat

Question 2What is the "agentic harness"?

AThe model itselfBThe permission systemCClaude Code — the tools, context management, and execution environment around the modelDThe terminal emulator

Question 3How do you interrupt Claude mid-task?

AClose the terminalBPress EscCType /stopDWait for the turn to finish

Further reading

• How Claude Code works — the agentic loop