A comprehensive toolkit for cutting OpenClaw token costs through context optimization, model routing, and heartbeat tuning. The context optimizer is the standout feature here, offering 50-80% savings by lazy-loading files based on prompt complexity instead of dumping everything upfront. Model routing enforces Haiku for casual chat and greetings so you never burn Opus tokens on "thanks." Heartbeat optimization spaces out polling intervals and respects quiet hours. All four core scripts run locally with no network calls, though the reference docs describe optional multi-provider setups if you want to route across OpenRouter or Together. Security audit is recent and the approach is solid. If you're running multiple agents or hitting rate limits, this pays for itself quickly.
npx skills add https://github.com/asif2bd/openclaw-token-optimizer --skill token-optimizer