Blog

AI Agent Glossary

Chain-of-thought (CoT)

Having the model write intermediate reasoning steps before its answer or action.

Chain-of-thought improves multi-step accuracy by giving the model space to plan. In agents it appears as the 'Thought' in ReAct, as hidden reasoning in reasoning models (o-series, Claude extended thinking), or as an explicit planning step. It increases tokens per step, which matters for long agent loops.

Source: Chain-of-Thought Prompting Elicits Reasoning in Large Language Models — Wei et al., 2022

See the numbers behind this term: AI Agent Statistics 2026.