weekly-cost-report

name: weekly-cost-report description: Use when a user asks for a cost report, model usage summary, token spend breakdown, API costs, or says "how much did I spend this week?", "show me my API costs", "what's my model spend?", "weekly cost summary", or "usage report". Parses OpenClaw JSONL session logs and aggregates per-model token usage and dollar cost for the last 7 days.

Weekly Cost Report

Overview

Parses OpenClaw session JSONL logs to produce a per-model cost breakdown for the past 7 days. Reports total spend, per-model costs with percentages, and daily spend trend.

Usage

python3 scripts/report.py

Output is printed to stdout. Pipe or forward to WhatsApp as needed.

Workflow

1. Run python3 scripts/report.py
2. Script scans all .jsonl files in $OPENCLAW_SESSIONS_DIR (default ~/.openclaw/agents/main/sessions)/
3. Filters to entries with timestamps in the last 7 days
4. Extracts message.usage.cost.total and message.model from each turn
5. Aggregates:
   • Total cost and turn count
   • Per-model cost and percentage of total
   • Per-day cost
6. Prints formatted WhatsApp-ready report

Output Format

📊 Weekly API Cost Report
Period: last 7 days

Total: $X.XXXX across N,NNN turns

By model:
  anthropic/claude-sonnet-4-6: $X.XXXX (XX.X%)
  ...

By day:
  2026-03-17: $X.XXXX
  ...

Scripts

• scripts/report.py — Standalone script. No arguments. Reads JSONL session logs, aggregates costs, prints report. Handles missing data gracefully (outputs "No API usage data found" if logs are empty or missing).

Requirements

• Python 3.8+ (stdlib only — json, os, glob, datetime)
• OpenClaw session logs at $OPENCLAW_SESSIONS_DIR (default ~/.openclaw/agents/main/sessions)/*.jsonl