Agent mode is the setting where the model stops being a chat partner and becomes a worker. Instead of answering your message once, it runs a loop: plan a step, take an action, look at the result, decide the next step, and keep going until the task is done or it needs you. The same underlying model powers both modes. Agent mode is about giving it the leash to act on its own.
What changes when you switch it on
In a plain chat, the model produces text and stops. You read it, you decide, you act. In agent mode the agent closes that loop itself:
- It chooses actions, calling tools to read files, run commands, and edit code without asking between each one.
- It works multi-step, chaining many actions toward a goal rather than answering a single question.
- It keeps going until it finishes or hits something it is not allowed to do.
That autonomy is exactly what makes it powerful and exactly what makes it risky.
Capability buys you trust problems
More independence means more can happen without you in the loop, good and bad. This is why agent mode leans hard on guardrails. The permission mode decides which actions still need your sign-off, and keeping a human in the loop for the consequential steps is what makes running autonomously safe rather than reckless.
Related terms
Agent
An agent is a language model wrapped in a loop that lets it call tools, read the results, and decide what to do next. The model supplies the judgement; the loop and the tools give it hands.
Read definition →Permission mode
Permission mode is the policy that decides which actions an agent can take on its own and which ones need your approval, ranging from ask-every-time to full auto. It trades safety for flow.
Read definition →Human in the loop
Human in the loop means keeping a person in the agent's decision path to approve, steer, or verify its work. It is the deliberate counterweight to full autonomy.
Read definition →