What problem does it solve?
It helps you run complex, multi-agent Claude Code collaborations without losing coordination, quality gates, or state—especially when tasks require research, implementation, review, debate, and clean shutdown.
Core Features & Use Cases
- Multi-role orchestration: Creates and coordinates teams with purpose-built roles (researchers, architect, implementers, reviewers, tester, synthesizer).
- 7-phase lifecycle execution: Classify → compose → get approval → spawn → coordinate & monitor → shutdown → cleanup & report.
- AA-MA structured memory integration: When an AA-MA task is active, it maps milestones to team tasks and keeps AA-MA files synchronized at the right times.
- Research debate for higher confidence: Implements a competing-hypotheses challenge round to reduce confirmation bias.
- Quality and error guardrails: Enforces design/code/test consistency gates and provides protocols for file conflicts, stuck teammates, partial completion, and retries.
- Graceful shutdown and final reporting: Drains active work, verifies results, deletes team artifacts, and summarizes outcomes to the user.
Quick Start
Use agent-teams when your task needs two or more coordinated agents that must communicate (for example, a research phase with competing hypotheses followed by coordinated implementation).
