cdt

Multi-agent development workflows are orchestrated by coordinating specialized agent teammates across plan, dev, full, and auto modes to execute end-to-end software changes. It assigns defined roles like Architect, Developer, and Reviewer to handle dependency-driven tasks and governance gates.

Managing agent teams for architecture planning is handled through the plan mode, which coordinates an Architect agent to produce plan artifacts and Architecture Decision Records (ADRs). This ensures structured governance and clear handoff material before development begins.

Automated testing and review gates are enforced within the full and auto modes by assigning Code-Tester, QA-Tester, and Reviewer agents. This ensures dependency-driven task execution validates code quality and governance before final handoff.

A multi-agent team with defined roles is needed for complex software tasks that benefit from parallel experimentation, structured governance, and cross-team messaging. It ensures end-to-end changes are properly planned, tested, and reviewed by specialized agents.

Multi-agent orchestration works without external dependencies, relying entirely on internal references to coordinate the agent teams. It generates its own governance prompts, plan artifacts, and cross-team messaging to drive the development workflow.

The modes represent staged collaboration phases: plan generates architecture and ADR artifacts, dev executes implementation, full applies testing and review gates, and auto enables end-to-end orchestration across all defined agent roles.