swarm-advanced

Initialize a swarm with swarm_init specifying a topology like mesh or hierarchical, then spawn agents with agent_spawn and orchestrate work via task_orchestrate. Alternatively, run npx claude-flow swarm with a task description and strategy flags from the CLI.

Use mesh for research and collaborative analysis, hierarchical for structured development workflows, star for centralized testing and validation, and ring for sequential pipeline processing. The choice depends on whether agents need peer communication or a coordinator.

Yes, parallel_execute runs independent tasks concurrently across spawned agents, while sequential strategies handle dependent tasks. You should monitor resource usage and implement error handling for parallel execution.

Yes, memory_usage stores data in namespaces with TTL values, and state_snapshot creates checkpoints you can restore later. Memory backups can be written to disk for cross-session persistence.

Coordination failures usually stem from wrong topology selection, missing memory configuration, or disabled monitoring. Verify the topology matches your workflow, check namespace usage, and enable swarm_monitor to observe agent health.