agent-team

Multi-agent parallel execution is coordinated by creating a named team and spawning specialized subagents to execute sub-tasks concurrently. It uses a shared task list and SendMessage to track progress and manage dependencies across backend, frontend, research, and testing domains.

Multi-agent orchestration decomposes complex multi-step tasks and assigns optimal built-in or project-specific subagents across domains like backend, frontend, research, and testing. It coordinates parallel execution, manages blocking dependencies, and synchronizes distributed software engineering work.

Yes, multi-agent team workflows support custom project-specific agents defined under the .claude/agents/ directory. These integrate alongside built-in agent types like Explore, Plan, Bash, code-simplifier, and general-purpose to handle specialized parallel sub-tasks.

Dependencies and communication between parallel subagents are managed using SendMessage and a shared task list. The TeamCreate and TaskCreate workflow provides blocking mechanisms and run_in_background options to coordinate execution and synchronize progress across agents.

Parallel multi-agent coordination is suited for complex multi-step tasks that benefit from distributed subagents across backend, frontend, research, and testing domains. It applies when tasks require concurrent execution, blocking dependency management, and specialized project context handling.