You'd reach for this when orchestrating multiple AI agents that need to coordinate work without a central controller. It acts as a message broker and task distributor, letting agents in a swarm claim jobs, relay status updates, and communicate asynchronously. The decentralized architecture means no single point of failure and agents can join or leave the swarm dynamically. Think of it as a lightweight pub/sub system specifically designed for multi-agent scenarios where you want parallel task execution without hardcoding dependencies between agents. The SSE transport keeps connections open for real-time message relay between swarm participants.
claude mcp add --transport sse io.github.evozim-swarm-coordinator https://swarm-coordinator-mcp.vercel.app/api/mcp