How Agents Share Information
Agents in a Flux workflow communicate through a shared context. When an agent completes its step, its output is stored and made available to every subsequent agent in the workflow.
Implicit Context Passing
By default, each agent receives the outputs of all previous steps as part of its context. This means a writer agent placed after a researcher will automatically have access to the research findings without any explicit configuration.
Structured Outputs
For more precise data passing, you can define output schemas on each step. Structured outputs let you specify exactly what data an agent should produce — such as a list of emails, a set of key-value findings, or a JSON object. Downstream agents and action steps can reference specific fields from these outputs.
Referencing Previous Steps
Action steps use a config field to reference data from earlier steps. For example, a send-email action might reference research_step_output.emails to get the list of recipients generated by a previous agent. The dot-notation resolver handles nested data access automatically.