ReAct agent

A ReAct agent works in a loop: it reasons about what it needs, takes an action to get it, observes the result, and reasons again. With Virgo, each action can be a query against the reasoning graph, so the agent grounds every step in your team's real context instead of guessing.

How it uses the graph

At each step the agent calls a retriever: the context retriever to understand how a system works, the causal retriever to see why it changed, or the learning retriever to recall a convention. Because retrieval is permission-aware, the agent only ever sees what its user can see.

When to use it

Reach for ReAct when a task is open-ended and the next move depends on what the last query returned, like investigating a bug or tracing a decision.

Related

• Agents: the model behind this pattern.
• Orchestrator agent and Scatter-gather agent.