cc-parallel-agents

To run parallel agent tasks, dispatch 2+ Agent tool calls in a single message to execute them concurrently. This pattern ensures isolated contexts for each agent, speeding up complex investigations like debugging multi-subsystem failures.

Debugging multi-subsystem failures with parallel agents involves dispatching independent investigative tasks concurrently in one message. This isolated context approach allows agents to investigate separate subsystems simultaneously without shared state interference.

Yes, you can use parallel agents for independent research tasks that require no shared state. The pattern dispatches multiple Agent tool calls in one message, ensuring each task runs in an isolated context concurrently.

After dispatching concurrent agents, you aggregate results using the provided aggregation workflow. Agents coordinate via the filesystem, allowing you to review and integrate the isolated findings into a final result.

Yes, parallel agent dispatch optionally supports worktree isolation for concurrent editing. This feature allows multiple agents to edit files simultaneously without overriding each other's changes during independent investigations.

You should avoid using parallel agents when tasks require shared state or have dependencies between them. The pattern is designed strictly for independent investigations where isolated contexts can operate without interfering with each other.