bmad-party-mode

To run a multi-agent roundtable discussion, spawn 2–4 BMAD subagents in parallel to independently answer the same question from distinct perspectives. This prevents shallow, converged outputs by ensuring each agent thinks and responds in its own voice.

Preventing convergent single-model roleplay requires orchestrating genuinely independent subagents. By spawning separate BMAD agents to discuss the same question, you eliminate shallow multiple-roles-in-one-model responses and get unabridged, diverse perspectives.

Include project context by loading the BMAD agent roster config and providing project-context.md. This resolves relevant voices and supplies the background needed for accurate cross-functional brainstorming and multi-perspective Q&A.

Yes, you can roleplay multiple agent personas without parallel subagents by using the --solo flag. This makes the model simulate all selected personas in a single pass instead of spawning independent subagents.

Align model speed and quality for trade-off discussions by using the --model flag during invocation. This lets you select a faster or deeper model to match the specific requirements of the roundtable round.

Avoid parallel execution for multi-agent orchestration if you need faster response times or lack subagent capabilities. Use the --solo flag instead to roleplay all selected personas within a single model pass.