Agent Mode
The mode where the assistant acts instead of suggesting.
A mode in which the tool takes actions — editing files, running commands, iterating — rather than proposing changes for you to apply. The opposite of a chat or suggestion mode.
Agent mode changes the review posture. You are no longer reading a diff before it exists; you are tracking a process and inspecting the result.
The change from suggesting to acting is a change in who is accountable for correctness, not just in convenience. When the tool edits directly, review becomes the gate rather than the delivery mechanism — which raises the importance of diffs, tests and reversible steps.
Small diffs survive review
An agent is asked to refactor a large module and returns one enormous change.
Nobody reviews five thousand lines properly. Constrain work to small verified increments — one behaviour per change, tests green between them — so review is possible and a bad step is cheap to revert.
- docsClaude Code overview ↗
A first-party description of the surfaces an agentic coding tool ships on, including parallel sessions and long-running tasks. Not evidence that it improves productivity.
- blogBuilding effective agents ↗
Draws the line between fixed workflows and agents that choose their own process, and names the core patterns: chaining, routing, parallelisation, orchestrator-workers, evaluator-optimizer.
Permission Mode · Autonomy level
How much the agent may do without asking.
Agent
A model given tools, a goal and permission to keep going on its own.
Automated Review
A model reviewing a model’s output before a human sees it.
AFK · Away from keyboard
Running without a person watching.