Token Manager

Universal LLM Token Manager - Monitor usage and provide cost-saving recommendations for Kimi, OpenAI, Anthropic, Gemini, and local models. Features scheduled monitoring, cross-session tracking, and proactive alerts.

Security Scan

The skill's code and runtime instructions align with a token-usage/cost-monitoring tool and its network/credential needs are proportionate, but the registry metadata omits declared environment variables and there are a few small inconsistencies you should review before installing.

Purpose & Capability

The code (manager/scheduler/session-tracker) implements token estimation, balance checks, scheduled alerts, and cross-session analytics as described. Network calls target LLM provider APIs (moonshot/OpenAI/Anthropic/Gemini/localhost) which is expected for a token manager. Pricing and provider tables in code match the stated purpose.

Instruction Scope

SKILL.md instructs the agent to read API keys from environment variables, store data under a local .data/ directory, register cron jobs, and edit openclaw.json to add tools. The instructions do not attempt to read unrelated system files or exfiltrate data. Note: SKILL.md and the scripts rely on environment variables (MOONSHOT_API_KEY, OPENAI_API_KEY, ANTHROPIC_API_KEY) even though the registry metadata declared no required env vars—this mismatch should be corrected/confirmed.

Install Mechanism

There is no install spec (no remote downloads). The package is provided as local JS scripts; no external install or archive extraction occurs. This minimizes install-time risk.

Credentials

The tool needs provider API keys to query balances/estimate tokens; those credentials are proportionate to its function. However, registry metadata lists no required env vars while SKILL.md and scheduler/manager scripts expect MOONSHOT_API_KEY (and optionally OPENAI_API_KEY / ANTHROPIC_API_KEY). This metadata omission is an inconsistency to fix before trusting automated runs.

Persistence & Privilege

The skill persists data only under its local .data/ directory (config.json, sessions, alerts). It does not request elevated system privileges or attempt to modify other skills or system-wide agent settings. always is false and autonomous invocation is enabled (the platform default).

Assessment

This skill appears to do what it says: it stores usage data locally, queries provider APIs with your API keys, and provides alerts/reports. Before installing: 1) Confirm and set the expected env vars (MOONSHOT_API_KEY, optionally OPENAI_API_KEY and ANTHROPIC_API_KEY) — the registry metadata currently omits them. 2) Review the scripts (they are included) and verify the provider endpoints and pricing entries match your expectations. 3) When registering cron jobs or editing openclaw.json, ensure you use the correct absolute path and do not run the scripts as root. 4) Understand the scripts will make outbound HTTPS/HTTP requests to the configured provider hosts using whatever API key you supply — only provide keys you trust to be used in this context. 5) If you do not trust the source, run the code in an isolated environment (container or VM) or manually audit/modify the code before enabling automated cron/tool registration.

Universal LLM Token Manager with proactive monitoring and analytics.

When to Use

Use this skill when you need to:

• Monitor LLM API token usage and costs
• Get cost-saving recommendations
• Set up automated balance alerts
• Track usage across multiple sessions
• Generate daily/weekly usage reports

Quick Start

cd /path/to/token-manager
export MOONSHOT_API_KEY="your-api-key"

# Generate report
node scripts/manager.js report 11000 146 42000 200000 off 9.26 moonshot kimi-k2.5

Core Features

1. Usage Monitoring

Real-time session analysis with cost-saving suggestions.

2. Scheduled Alerts (P0)

Automatic balance monitoring with proactive notifications.

3. Built-in Tool Integration (P1)

Register as OpenClaw tool for seamless usage.

4. Cross-Session Analytics (P2)

Track spending patterns and generate reports.

Supported Providers

Provider	Balance Query	Token Estimate	Pricing
Kimi/Moonshot	✅ API	✅ API	¥12/1M tokens
OpenAI	❌ Console	❌ Approx	USD/1M tokens
Anthropic/Claude	❌ Console	❌ Approx	USD/1M tokens
Google/Gemini	❌ Console	❌ Approx	USD/1M tokens
Ollama/Local	N/A Free	N/A	FREE

Cost-Saving Recommendations

Context Management

Scenario	Recommendation	Action
Context > 80%	🚨 Critical: Must compact immediately	/compact
Context > 50%	📊 Suggest: Consider compacting	/compact
Session > 50k tokens	⚠️ Warning: Split tasks now	/spawn
Session > 20k tokens	💡 Tip: Use sub-agents for large tasks	/spawn

Reasoning Optimization

Scenario	Recommendation	Action
Reasoning ON + small task (<5k tokens)	💡 Can disable to save 20-30%	/thinking off
Reasoning ON + complex task	✅ Keep on for quality	Keep

Provider-Specific Tips

Scenario	Recommendation
Balance < ¥5	🚨 Enable save mode, avoid large tasks
Using GPT-4	💡 Consider GPT-4o-mini for 10x savings
Using Claude Opus	💡 Consider Claude Sonnet for 5x savings
Running Ollama	🎉 Free! No API costs

Commands

Manager (Core)

node scripts/manager.js report <tokensIn> <tokensOut> <contextUsed> <contextMax> <thinking> [balance] [provider] [model] [apiKey]
node scripts/manager.js balance [provider] [apiKey]
node scripts/manager.js estimate <provider> <inputTokens> <outputTokens> [model]
node scripts/manager.js providers
node scripts/manager.js history

Scheduler (P0 - Cron Alerts)

# Check balance and alert if below threshold
node scripts/scheduler.js check <provider> <threshold>

# View alert statistics
node scripts/scheduler.js stats

Session Tracker (P2 - Analytics)

# Record session for tracking
node scripts/session-tracker.js record <provider> <model> <tokensIn> <tokensOut> <cost> [currency]

# Generate reports
node scripts/session-tracker.js daily [date]
node scripts/session-tracker.js weekly
node scripts/session-tracker.js recommend

P0: Scheduled Monitoring & Alerts

Setup automatic balance monitoring with cron jobs.

Setup Cron Job

# Check every hour, alert if below ¥5
openclaw cron add \
  --name "token-balance-check" \
  --schedule "0 * * * *" \
  --command "cd /path/to/token-manager && node scripts/scheduler.js check moonshot 5"

Alert Rules

Condition	Action	Cooldown
Balance < threshold	Send alert	1 hour
Balance < ¥1	Send urgent alert	30 min
3 alerts in 24h	Suggest adding funds	-

Alert Output

When triggered, outputs JSON:

{
  "alert": true,
  "balance": 3.50,
  "threshold": 5,
  "messages": {
    "en": "🚨 [URGENT] Token Manager Alert...",
    "cn": "🚨 [紧急] Token 管家提醒..."
  }
}

P1: Tool Integration

Register as OpenClaw tool for direct usage.

Tool Configuration

Add to openclaw.json:

{
  "tools": {
    "token_status": {
      "command": "cd /path/to/token-manager && node scripts/manager.js report",
      "description": "Check current token usage and costs"
    },
    "token_balance": {
      "command": "cd /path/to/token-manager && node scripts/manager.js balance",
      "description": "Query account balance"
    }
  }
}

Usage After Registration

# Direct commands
openclaw tool token_status 11000 146 42000 200000 off 9.26 moonshot
openclaw tool token_balance moonshot

P2: Cross-Session Tracking

Track usage patterns across multiple sessions.

Recording Sessions

Automatically or manually record each session:

node scripts/session-tracker.js record moonshot kimi-k2.5 5000 500 0.06 CNY

Daily Report

node scripts/session-tracker.js daily
# Output: Total tokens, cost, session count, provider breakdown

Weekly Report

node scripts/session-tracker.js weekly
# Output: 7-day summary with trend analysis

Smart Recommendations

node scripts/session-tracker.js recommend
# Analyzes patterns and suggests optimizations

Environment Variables

• MOONSHOT_API_KEY - Kimi/Moonshot API key
• OPENAI_API_KEY - OpenAI API key (optional)
• ANTHROPIC_API_KEY - Anthropic API key (optional)

Security

• API keys read from environment variables only
• All data stored locally in .data/ directory
• No data uploaded to third-party servers
• Network requests only to official LLM APIs
• Alert state persisted locally with cooldown logic

Pricing Reference

Kimi/Moonshot

• K2.5: ¥12 / 1M tokens

OpenAI

• GPT-4o: $2.5 / $10 per 1M
• GPT-4o-mini: $0.15 / $0.6 per 1M
• GPT-3.5-turbo: $0.5 / $1.5 per 1M

Anthropic

• Claude 3.5 Sonnet: $3 / $15 per 1M
• Claude 3 Opus: $15 / $75 per 1M
• Claude 3 Haiku: $0.25 / $1.25 per 1M

Google Gemini

• Gemini 1.5 Pro: $3.5 / $10.5 per 1M
• Gemini 1.5 Flash: $0.35 / $1.05 per 1M

Ollama

• Local execution: FREE

---

---

Token 管家

通用 LLM Token 管理工具，支持主动监控和数据分析。

使用场景

在以下情况使用此 skill：

• 监控 LLM API token 使用和费用
• 获取省钱优化建议
• 设置自动余额提醒
• 追踪多会话使用模式
• 生成每日/每周使用报告

快速开始

cd /path/to/token-manager
export MOONSHOT_API_KEY="your-api-key"

# 生成报告
node scripts/manager.js report 11000 146 42000 200000 off 9.26 moonshot kimi-k2.5

核心功能

1. 使用监控

实时会话分析，提供省钱建议。

2. 定时提醒 (P0)

自动余额监控，主动通知。

3. 工具集成 (P1)

注册为 OpenClaw 工具，无缝使用。

4. 跨会话分析 (P2)

追踪消费模式，生成报告。

支持的提供商

提供商	余额查询	Token 估算	价格
Kimi/Moonshot	✅ API	✅ API	¥12/百万
OpenAI	❌ 控制台	❌ 估算	USD/百万
Anthropic/Claude	❌ 控制台	❌ 估算	USD/百万
Google/Gemini	❌ 控制台	❌ 估算	USD/百万
Ollama/本地	N/A 免费	N/A	免费

省钱优化建议

上下文管理

场景	建议	操作
上下文 > 80%	🚨 紧急：必须立即压缩	/compact
上下文 > 50%	📊 建议：适时压缩	/compact
会话 > 50k tokens	⚠️ 警告：立即拆分任务	/spawn
会话 > 20k tokens	💡 提示：大任务使用子代理	/spawn

推理优化

场景	建议	操作
Reasoning 开启 + 小任务 (<5k tokens)	💡 可关闭节省 20-30%	/thinking off
Reasoning 开启 + 复杂任务	✅ 保持开启确保质量	保持

提供商特定建议

场景	建议
余额 < ¥5	🚨 开启省钱模式，避免大任务
使用 GPT-4	💡 考虑 GPT-4o-mini 省 10 倍
使用 Claude Opus	💡 考虑 Claude Sonnet 省 5 倍
运行 Ollama	🎉 免费！无 API 费用

命令

管理器（核心）

node scripts/manager.js report <输入tokens> <输出tokens> <上下文已用> <上下文上限> <推理状态> [余额] [提供商] [模型] [apiKey]
node scripts/manager.js balance [提供商] [apiKey]
node scripts/manager.js estimate <提供商> <输入tokens> <输出tokens> [模型]
node scripts/manager.js providers
node scripts/manager.js history

调度器 (P0 - 定时提醒)

# 检查余额，低于阈值时提醒
node scripts/scheduler.js check <提供商> <阈值>

# 查看提醒统计
node scripts/scheduler.js stats

会话追踪器 (P2 - 分析)

# 记录会话
node scripts/session-tracker.js record <提供商> <模型> <输入tokens> <输出tokens> <费用> [货币]

# 生成报告
node scripts/session-tracker.js daily [日期]
node scripts/session-tracker.js weekly
node scripts/session-tracker.js recommend

P0: 定时监控与提醒

使用 cron 设置自动余额监控。

设置定时任务

# 每小时检查，低于 ¥5 时提醒
openclaw cron add \
  --name "token-balance-check" \
  --schedule "0 * * * *" \
  --command "cd /path/to/token-manager && node scripts/scheduler.js check moonshot 5"

提醒规则

条件	动作	冷却时间
余额 < 阈值	发送提醒	1 小时
余额 < ¥1	发送紧急提醒	30 分钟
24 小时内 3 次提醒	建议充值	-

提醒输出

触发时输出 JSON：

{
  "alert": true,
  "balance": 3.50,
  "threshold": 5,
  "messages": {
    "en": "🚨 [URGENT] Token Manager Alert...",
    "cn": "🚨 [紧急] Token 管家提醒..."
  }
}

P1: 工具集成

注册为 OpenClaw 工具直接使用。

工具配置

添加到 openclaw.json：

{
  "tools": {
    "token_status": {
      "command": "cd /path/to/token-manager && node scripts/manager.js report",
      "description": "Check current token usage and costs"
    },
    "token_balance": {
      "command": "cd /path/to/token-manager && node scripts/manager.js balance",
      "description": "Query account balance"
    }
  }
}

注册后使用

# 直接命令
openclaw tool token_status 11000 146 42000 200000 off 9.26 moonshot
openclaw tool token_balance moonshot

P2: 跨会话追踪

追踪多会话使用模式。

记录会话

自动或手动记录每个会话：

node scripts/session-tracker.js record moonshot kimi-k2.5 5000 500 0.06 CNY

每日报告

node scripts/session-tracker.js daily
# 输出：总 token、费用、会话数、提供商分布

每周报告

node scripts/session-tracker.js weekly
# 输出：7 天摘要及趋势分析

智能建议

node scripts/session-tracker.js recommend
# 分析模式并提供优化建议

环境变量

• MOONSHOT_API_KEY - Kimi/Moonshot API 密钥
• OPENAI_API_KEY - OpenAI API 密钥（可选）
• ANTHROPIC_API_KEY - Anthropic API 密钥（可选）

安全说明

• API 密钥仅从环境变量读取
• 所有数据本地存储在 .data/ 目录
• 无数据上传到第三方服务器
• 网络请求仅访问官方 LLM API
• 提醒状态本地持久化，带冷却逻辑

价格参考

Kimi/Moonshot

• K2.5: ¥12 / 百万 tokens

OpenAI

• GPT-4o: $2.5 / $10 每百万
• GPT-4o-mini: $0.15 / $0.6 每百万
• GPT-3.5-turbo: $0.5 / $1.5 每百万

Anthropic

• Claude 3.5 Sonnet: $3 / $15 每百万
• Claude 3 Opus: $15 / $75 每百万
• Claude 3 Haiku: $0.25 / $1.25 每百万

Google Gemini

• Gemini 1.5 Pro: $3.5 / $10.5 每百万
• Gemini 1.5 Flash: $0.35 / $1.05 每百万

Ollama

• 本地运行：免费