token用量实时监控

Monitor and analyze OpenClaw token usage across sessions. Shows models used, token counts (input/output/cache), daily trends, and hourly breakdowns. Use when the user asks about token usage, API consumption, cost tracking, model usage stats, token trends, or how many tokens have been used. Triggers: token usage, 用量, token统计, 消耗, cost, 花费, model usage, API usage.

Token Usage Monitor

Analyze OpenClaw session token usage from JSONL session files.

Quick Usage

Run the analysis script:

python3 ~/.openclaw/workspace/skills/token-usage-monitor/scripts/analyze_tokens.py

Options

| Flag | Default | Description | |------|---------|-------------| | --days N | 7 | Include last N days (0 = all time) | | --json | false | Output as JSON for programmatic use | | --top N | 10 | Show top N models | | --base-dir PATH | ~/.openclaw/agents | Custom agents directory |

Examples

# Last 7 days (default)
python3 scripts/analyze_tokens.py

# Last 30 days
python3 scripts/analyze_tokens.py --days 30

# All time, JSON output
python3 scripts/analyze_tokens.py --days 0 --json

# Top 5 models
python3 scripts/analyze_tokens.py --top 5

What It Shows

Overview — Total API calls, input/output/cache tokens, average per call
By Model — Token breakdown per model with percentage share
Daily Trend — Day-by-day token consumption with visual bars
Hourly Breakdown — Usage distribution for the latest active day

Data Source

Reads from ~/.openclaw/agents/*/sessions/*.jsonl — all agent session files. Each assistant message includes a usage object with input, output, cacheRead, cacheWrite, totalTokens, and cost fields.

Notes

Cost is $0.00 if the provider doesn't return pricing (e.g., some local/custom providers)
The script aggregates across ALL agents (main, sub-agents, etc.)
Timestamps are in UTC; daily grouping uses UTC dates