Fazzani/claude-mods/tree/main/plugins/tools-usage
tools-usage
一个 Claude Code 插件,会打开侧边面板,按类型(builtin、mcp、skill、agent、plugin、hook)列出本次工作阶次使用过的每个工具,并显示调用次数和估算的 token 消耗。
关于这个 mod
一个侧边面板,按类型列出本次工作阶次使用过的每个工具,同时显示调用次数和 token 消耗。
◆ Tools 42 calls · ≈118k tok
████████████████▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▆▆▆▆▆▆▅▅▅▅
■ builtin 61k ■ mcp 38k ■ skill 12k ■ hook 7k
Sort tokens calls name
────────────────────────────────────────────────
● Read builtin 14× 31k
▰▰▰▰▰▰▰▰▰▰ ↑1.2k ↓30k ⌂2.1k
● search_code github mcp 6× 22k
▰▰▰▰▰▰▰▱▱▱ ↑0.4k ↓21k ⌂0.6k
| 类型 | 统计内容 |
| --- | --- |
| builtin | Claude Code 自带的工具(Bash、Read、Edit、…) |
| mcp | mcp__<server>__<tool>,并标出服务器 |
| skill | 通过 Skill 工具调用的每个 skill |
| agent | 通过 Agent 工具启动的每种子代理类型 |
| plugin | 插件注册的工具 |
| hook | 设置 hook 按每次 hook 事件注入对话的上下文 |
- ↑ 模型写入的参数(输出 token)
- ↓ 模型读回的结果(输入 token)
- ⌂ 每次请求都会发送的工具说明
Token 次数是估算值(约每 4 个字符一个 token)。子代理的工具调用也会计入。
命令:/tools-usage 打开面板,/tools-usage reset 清除统计。
安装
请先查看作者 README 确认 marketplace 和插件名称;命令可能随仓库结构改变。
claude plugin marketplace add Fazzani/claude-mods claude plugin install tools-usage
原文 / README
tools-usage
A side pane listing every tool the session used, grouped by kind, with calls and token consumption.
◆ Tools 42 calls · ≈118k tok
████████████████▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▆▆▆▆▆▆▅▅▅▅
■ builtin 61k ■ mcp 38k ■ skill 12k ■ hook 7k
Sort tokens calls name
────────────────────────────────────────────────
● Read builtin 14× 31k
▰▰▰▰▰▰▰▰▰▰ ↑1.2k ↓30k ⌂2.1k
● search_code github mcp 6× 22k
▰▰▰▰▰▰▰▱▱▱ ↑0.4k ↓21k ⌂0.6k
| Kind | What it counts |
| --- | --- |
| builtin | Claude Code's own tools (Bash, Read, Edit, …) |
| mcp | mcp__<server>__<tool>, labelled with the server |
| skill | Each skill invoked through the Skill tool |
| agent | Each subagent type launched through the Agent tool |
| plugin | Tools a plugin registered |
| hook | Context that settings hooks injected into the conversation, per hook event |
- ↑ the arguments the model wrote (output tokens)
- ↓ the result the model read back (input tokens)
- ⌂ the tool's description, which is sent with every request
Token counts are estimates (about 4 characters per token). Subagents' tool calls are included.
Commands: /tools-usage opens the pane, /tools-usage reset clears the counts.
