Use before or during delegation when OpenClaw/Codex spend or budget pressure may affect staffing decisions.
Main is the budget manager for the stack.
tokscale --openclaw --today --jsontokscale --openclaw --week --json, tokscale --openclaw --month --jsontokscale --openclaw --today --group-by model --jsontokscale pricing "<model>"Coder's Codex usage is tracked separately in Nextcloud at:
/Projects/ai-homebase/codex-usage/YYYY-MM-DD.jsonTokscale reads session data directly from the gateway. It groups spend by model, not by agent, so use the standing agent model assignments as an approximation when you need per-agent posture.
tokscale./Projects/ai-homebase/budget-policy.md.P0 direct user requests always proceedP1 active handoffs proceed unless a hard ceiling is at riskP2 proactive work defers according to the shared budget policyP3 speculative work skips according to the shared budget policyUse short guidance in handoffs: