Fundamentals

An agent describes a conversational structure that breaks interactions down into Stages and Transitions.

This combination of Stages and Transitions allows a full conversational structure to be described. There are no assumptions made about the type of structure i.e. it may be linear, branching, circular etc.

The LLM can then choose which paths to navigate within that structure based on the interaction with the user and the prompting provided by the author.

The Agent directs the LLM to perform tasks using Prompting and can also instruct the agent to show to or collect data via Variables.

Retrieving extra context and communicating Agent results with the outside world are achieved using Integrations.

Taken together, this conversational structure enables and encourages the core expressiveness and intelligence of the LLM in service of your use case while still constraining it toward a predictable outcome.