ReAct (reason + act)
An agent pattern that interleaves reasoning steps with tool actions in a loop, so the agent thinks, acts, observes the result, and reasons again toward the goal.
In more detail
ReAct describes a loop in which an agent alternates between reasoning about what to do and acting on it. It reasons, calls a tool, observes the outcome, reasons about the new state, and acts again, continuing until the task is complete. The interleaving lets the agent adapt to what it actually finds rather than committing to a plan made blind.
The pattern is powerful because it grounds the agent's reasoning in real observations at each step, reducing the chance of drifting into a plausible but wrong plan. Its cost is that every loop is more inference and more tool calls, so the loop needs bounds to keep it from running long or repeating itself.
Where this shows up at Ceven
Ceven's agents work in this reason-and-act spirit: a step can gather information, evaluate it, and decide the next action rather than following a blind script. The loop runs inside a bounded workflow with approval gates and an audit trail, so the adaptiveness of the pattern comes without giving up oversight of what the agent actually did.