Here, you'll find context on single-step tool use and multi-step tool use.

Single-step is sometimes referred to as "function calling" because it uses functions to call external tools that augment the capabilities of large language models. Multi-step tool use is sometimes referred to as "agents," because it forms the core of many workflows relying on agents.

You'll also find additional documentation on the various types of parameters offered by Cohere's tool use functionality, as well as a guide on when to use single-step v.s. multi-step capabilities.