Show HN: Lessons From Running Claude Code Swarms at Scale

Hacker News·medium signal

A practitioner posted lessons learned from operating large fleets of Claude Code agents in parallel ('swarms'), covering coordination overhead, token economics, and failure isolation. The write-up is directly relevant to builders orchestrating multi-agent coding workflows rather than single sessions. It complements a separate HN post on how bad MCP design can cost an agent up to 5x more tokens.