Orchestration · Software component

Conversational Agent Coordinator

Software componentOrchestrationOrchestration & Toolsarc:GroupChatManager

A workflow orchestrator in which autonomous agents, each keeping its own message history and auto-reply behaviour, advance work through natural-language message exchange rather than predefined control flow.

Responsibility. Advances a multi-agent task through turn-based conversational message exchange.

Also known as: Conversation-driven multi-agent coordination, Agent group chat, GroupChat, GroupChatManager, Conversational coordinator, Speaker selector, Conversational Agent Coordinator

Variant of Multi-Agent Coordinator abstract

When to choose. Choose when collaboration benefits from flexible dialogue (agents challenging outputs, requesting clarification, negotiating), for exploratory workflows, prototyping and human-overseen scenarios where readable transcripts aid transparency; avoid where determinism and structured audit trails are required.

Direct neighbourhood (hover for relationship types)

Relationships

invokes dependency

writes dependency

Shared Message History Ch2.4

is constrained by control

Iteration Limit Policy Ch2.4

orchestrates control

alternative to variability

Design guidance

SHOULD NOT be used where regulation requires identical decisions for identical inputs, because conversation order and content vary across runs even at temperature 0.
SHOULD add custom structured state management when workflow state goes beyond conversational context.
SHOULD NOT be used where compliance, auditability or reproducibility require identical inputs to yield identical execution paths, because conversation flow is non-deterministic.
SHOULD bound conversations with a maximum round limit and explicit termination conditions.
SHOULD pair with human validation of final outputs when conversational flexibility is used.

Classification

Patterns: Conversation-driven collaborationEmergent coordination through dialogueConversation-driven coordinationLLM-based speaker selectionTurn-takingCritique-revise loop
Technologies: AutoGenAutoGen GroupChatAutoGen ConversableAgent
Quality attributes: Flexibility (ISO/IEC 25010)Transparency and accountability (NIST AI RMF: accountable and transparent)
Risks mitigated: Rigid workflows unable to handle unpredictable agent involvement

Sources

Ch2.1: T. Nguyen, "Framework Landscape and Selection," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 2.1. ISBN: 9798244538229.
Ch2.4: T. Nguyen, "Multi-Agent Frameworks," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 2.4. ISBN: 9798244538229.