team-brainstorm

Multi-agent brainstorming works by running coordinated role workers that generate ideas, apply critique loops, synthesize themes, and score outputs to converge on well-evaluated proposals. This pipeline challenges weak assumptions to prevent settling on first drafts.

Run a structured brainstorming session by providing a topic to trigger the multi-agent pipeline. It spawns role workers from specifications, manages task dependencies, enforces generator-critic revision limits, and writes ideas, critiques, and rankings into per-phase directories.

Yes, multi-agent critique loops suit team workflows requiring structured idea exploration across multiple angles. The pipeline scales with topic complexity by offering Quick, Deep, and Full modes, including up to two rounds of generator-critic revision.

The best way to synthesize ideas is using a generator-critic pipeline that automatically challenges weak assumptions. A challenger agent critiques initial concepts, forcing revisions before a synthesis agent converges the refined outputs into a set of evaluated proposals.

No external dependencies are required to run session-based ideation and scoring. The workflow operates independently, utilizing a built-in session coordinator to spawn role workers, manage task dependencies, and persist structured artifacts into a consistent workflow directory.

Automated idea generation limits critique loops to a maximum of two generator-critic revision rounds. While it effectively challenges assumptions and synthesizes themes, extremely nuanced or highly specialized domain logic may require manual review beyond the automated evaluation artifacts.