Agents

Virgo gives AI tools company-grounded reasoning. Through it, an agent can retrieve the right context, understand prior decisions, and preserve what it learns for the engineers and agents that come next.

How agents use the graph

Agents read the graph through retrievers: they ask for the context behind a feature, the reasoning behind a decision, or the conventions a team already follows. Because retrieval is permission-aware, an agent only ever sees what its user can see.

Agents also write back. Useful learnings from an agent workflow are captured into the graph so they become available to future agents, new engineers, and later workflows. Context that took one agent effort to establish is not rediscovered from scratch by the next.

Agent patterns

Virgo supports several agent patterns, each suited to a different shape of problem:

• ReAct agent: reason and act in a loop against the graph.
• Orchestrator agent: coordinate sub-agents over a larger task.
• Scatter-gather agent: fan out across the graph and combine the results.

Related

• Retrievers: how agents read the graph.
• Custom retrievers: tailor what agents can pull.