Claude Code Token Monitor

中文

一款 VSCode 扩展，在状态栏实时显示 Claude Code 对话的 token 用量、费用和思考时间。

实时状态栏监控 — 显示会话标题、本轮 AI 耗时、本轮 token/费用、累计用量。悬停 Tooltip 查看完整分解：总输入 / 缓存命中 / 未命中 / 输出 / 命中率 / 思考时间
本轮 AI 耗时 — 统计当前对话轮次中 AI 实际工作的总时间，自动排除等待工具授权和用户输入的空闲时间
时间范围筛选 — 当前会话 / 每天 / 每周 / 每月 / 每年 / 全部，按逐条消息时间戳精确聚合
会话选择器 — 浏览所有过往会话，点击查看完整消息记录和分模型统计
按模型分组 — 一键切换，按 AI 模型拆分费用和 token 统计
多供应商定价 — 内置 11 家供应商 28+ 模型：DeepSeek · Anthropic · OpenAI · Gemini · Qwen · Kimi · GLM · Doubao · Ernie · Grok · Mistral
自动更新定价与汇率 — 汇率每日从 open.er-api.com 获取（免费，无需 key），模型定价每月通过 GitHub Actions 自动抓取更新
11 种语言 — en / zh-CN / zh-TW / ja / ko / es / ar / pt / de / fr / ru，切换语言时 webview 按钮实时更新无需重载
7 种货币 — CNY / USD / EUR / JPY / KRW / GBP + 自动检测
预算告警 — 累计费用超过阈值自动提示
自定义模型 — 命令面板引导添加，无需手写 JSON
零依赖，纯本地 — 不上传任何数据，除可选定价更新外完全离线

Ctrl+, → 搜索 claudeTokenMonitor：

A VSCode extension that displays real-time token usage, cost, and thinking time for Claude Code sessions in the status bar.

Real-time status bar — session title, turn AI time, per-turn tokens/cost, cumulative total. Hover tooltip with full breakdown: total input, cache hit/miss, output, hit rate, think time
Turn AI time — tracks total AI working time within the current conversation turn, excluding idle time waiting for tool approvals or user input
Time range filters — Current / Daily / Weekly / Monthly / Yearly / All, aggregated by per-message timestamps for accurate period stats
Session selector — browse all past sessions with full message history and per-model stats
By-model grouping — toggle to see costs split by AI model
Multi-provider pricing — 28+ models across 11 providers: DeepSeek · Anthropic · OpenAI · Gemini · Qwen · Kimi · GLM · Doubao · Ernie · Grok · Mistral
Auto-updating pricing & rates — exchange rates refreshed daily from open.er-api.com (free, no API key). Model pricing updated monthly via GitHub Actions
11 languages — en / zh-CN / zh-TW / ja / ko / es / ar / pt / de / fr / ru. Webview labels update instantly on language change without reload
7 currencies — CNY / USD / EUR / JPY / KRW / GBP + auto-detect
Budget warning — alerts when cumulative cost exceeds threshold
Custom models — guided quick-add via command palette, no manual JSON required
Zero dependencies, fully local — no telemetry, no data uploads, offline except optional pricing updates

Mode	Example
Full	`📊 Title… \\| AI time 3.2s \\| 85K ¥0.012 \\| Cumulative 1.2M ¥0.19`
Full + Model	`📊 Title… \\| deepseek-v4-pro \\| AI time 3.2s \\| 85K ¥0.012 \\| Cumulative 1.2M ¥0.19`
Compact	`📊 Title… \\| 1.2M ¥0.19`

Ctrl+, → search claudeTokenMonitor:

Command	Description
`Show Token Usage Detail`	Open token monitoring detail panel
`Show Last Message Cost`	Show cost of the last message
`Open Token Monitor Settings`	Open extension settings
`Refresh Token Data`	Manually refresh token data
`Add Custom Model Pricing`	Add custom model pricing
`Manage Custom Models`	Manage custom model entries
`Force Refresh Pricing Data`	Force refresh pricing data from remote

MIT