swarm-advanced

Orchestrate multi-agent systems by initializing a swarm topology (mesh, hierarchical, star, or ring), spawning specialized agents by type, and coordinating tasks using parallel or sequential strategies. The swarm manages distributed work across research, development, and testing phases with adaptive load balancing and knowledge graph integration for tracking findings.

Four topologies structure agent coordination: mesh enables peer-to-peer communication, hierarchical uses layered delegation, star routes through a central node, and ring chains agents sequentially. Four strategies optimize task distribution: adaptive adjusts dynamically, balanced spreads evenly, specialized assigns by expertise, and parallel executes concurrently.

Yes. Deploy a hierarchical development swarm where an architect designs the API structure, multiple coder agents implement endpoints in parallel, and tester agents validate unit, integration, and end-to-end tests concurrently. Swarm status and workflow execution APIs monitor progress and findings.

Initialize a swarm using mcp__claude-flow__swarm_init() with topology and maxAgents parameters. Spawn specialized agents via mcp__claude-flow__agent_spawn() by type (researcher, developer, tester) and name. Orchestrate tasks with mcp__claude-flow__task_orchestrate() specifying task details and execution strategy.

Swarm complexity grows with topology size and agent count; mesh topologies require more coordination overhead than star. Knowledge graph integration and memory tracking scale with findings volume. Performance depends on MCP tool latency and CLI command responsiveness across all participating agents.

Advanced implementation depth assumes familiarity with distributed system concepts and multi-agent coordination, but the Skill provides predefined topologies and strategies to simplify orchestration. Starting with mesh topology and parallel strategy suits teams new to swarm workflows.