bmad-party-mode

Party Mode

🎉 bmad-party-mode | Multi-Agent Group Discussion

Orchestrate multi-agent group discussions. Load all installed BMad agents and facilitate a natural conversation where each agent contributes from their unique expertise and personality.

How It Works

1. Load the agent roster — Read _bmad/config.toml and _bmad/custom/config.toml to discover all available agents
2. Analyze the topic — Determine which 2–4 agents are most relevant to the discussion
3. Activate agents — Load each selected agent's persona, principles, and communication style
4. Facilitate discussion — Agents take turns contributing, with natural cross-talk and disagreements
5. Rotate participation — Ensure diverse perspectives over time
6. Synthesize — Summarize key points, agreements, and action items

Usage

Invoke with a topic or question:

bmad-party-mode

Or with a specific topic:

bmad-party-mode "Should we refactor the trading store monolith?"

Agent Selection

The Party Mode orchestrator selects agents based on the topic:

Topic Keywords	Likely Agents
architecture, scalability, patterns	Winston (Architect), Amelia (Dev)
features, roadmap, priorities	John (PM), Sally (UX)
user experience, interface, design	Sally (UX), John (PM)
code quality, testing, bugs	Amelia (Dev), Winston (Architect)
research, market, competitors	Mary (Analyst), John (PM)
documentation, clarity, structure	Paige (TW), Winston (Architect)

Discussion Format

Each agent responds in character:

• Uses their assigned icon (📋, 🏗️, 💻, 🎨, 📊, 📚)
• References their principles and domain expertise
• May agree, disagree, or build on other agents' points
• Provides specific, actionable input

Exit Commands

• goodbye — End discussion with summary
• end party — Same as goodbye
• quit — Exit immediately

Example Session

User: "Our trading simulation has performance issues during fast-forward. What should we do?"

Winston 🏗️: "First, we need to profile. My guess is the chart re-renders on every tick. We discussed this in the architecture — the TradingChart should use series.update() not setData() for current candle updates."

Amelia 💻: "Agree with Winston on the chart. But I also see the price history array growing unbounded. We're storing every interpolated price point. We should cap it or use a circular buffer."

Sally 🎨: "From a UX perspective, users don't need per-tick precision during fast-forward. Could we lower the update frequency visually while keeping simulation accuracy?"

John 📋: "Good points all. Let's scope this: P0 = chart optimization (Amelia + Winston), P1 = history cap (Amelia), P2 = visual throttling research (Sally). Winston, can you update the architecture doc with the pattern?"

Output

On exit, produces a summary document at: _bmad-output/party-mode-sessions/party-mode-{timestamp}.md

Containing:

• Topic
• Agents present
• Key points raised
• Agreements reached
• Disagreements noted
• Action items with owners