Detect and reduce wasteful AI token usage by warning on large outputs, verbose logs, and repeated history in MCP-compatible clients.