Feature request: "Token savings mode" - reduce overall session token usage by default

Claude Code has a Concise output style for shorter prose, but there's no mode that targets overall token consumption - reduced tool-call narration, tighter context handling, minimized redundant restating of plan/context, avoiding re-reading files already in context, etc. Would like an opt-in (or default) "token savings mode" that optimizes for total tokens used per session, not just response brevity.