chinese-char-counter-mcp
You can count and extract Chinese characters from text via three read-only MCP tools.
Count Chinese characters in a single text, excluding punctuation, spaces, letters, digits, emoji, etc.
Get total characters, Chinese ratio, and a breakdown of Chinese, letters, digits, punctuation, spaces, and other.
Batch-count up to 500 texts with per-item results and a total Chinese count.
Extract only the Chinese characters from mixed text.
Use all tools safely since they are read-only and idempotent.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@chinese-char-counter-mcp 数一下这段文案有几个中文字:新品上市,全场五折!"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
chinese-char-counter-mcp
一个用 Python 实现的 MCP 服务(Model Context Protocol Server,STDIO 传输), 用于统计一段文本里的中文字数——标点、空格、英文字母、数字、emoji 等一律不计入。
传输方式:STDIO(本地进程,无网络、无 API Key、无环境变量)
运行时依赖:Python >= 3.10、
mcp>=1.0.0,<2已通过 ModelScope MCP 广场托管部署所需的配置形态(
command为uvx,包发布到 PyPI)
这个服务解决什么问题
大模型写中文文案时,"字数"常常对不上:把标点、空格、英文单词都算进去,或者把 emoji 也算成字。本服务给出一个确定的答案:只数中文字符,并把标点/字母/数字/空白 等分项一并返回,方便直接用于文案校验、作业字数检查、标题长度限制等场景。
Related MCP server: text-count-mcp-server
客户端配置
把下面这段配置加入任意支持 MCP 的客户端(Claude Desktop、Cursor、Cherry Studio、通义灵码、 ModelScope MCP 实验场等):
{
"mcpServers": {
"chinese-char-counter": {
"command": "npx",
"args": ["-y", "chinese-char-counter-mcp@latest"]
}
}
}说明:
该配置用 npm 上的零依赖 Node 实现(
npx会自动下载并运行,无需手动安装)。本服务不需要任何环境变量,因此配置里没有
env字段。想用 Python 实现:
pip install chinese-char-counter-mcp安装后,把配置里的command改成uvx、args改成该包名(即uvx chinese-char-counter-mcp@latest)即可;两种实现的工具名、参数、返回结构完全一致。command用npx或uvx是 ModelScope(魔搭)MCP 广场托管部署检测的硬要求,其他写法不会被托管。
安装
# 方式一:pip 安装后直接启动(控制台命令)
pip install chinese-char-counter-mcp
chinese-char-counter-mcp
# 方式二:模块方式启动
python -m chinese_char_counter_mcp
# 方式三:不安装,临时运行(需要 uv)
uvx chinese-char-counter-mcp@latestSTDIO 服务启动后不打印任何内容、等待客户端的 JSON-RPC 请求,这是正常现象。
工具
1. count_chinese_characters
统计单条文本的中文字数。
参数 | 类型 | 必填 | 说明 |
| string | 是 | 待统计文本,长度上限 200000 个字符 |
返回字段:
字段 | 类型 | 说明 |
| integer | 中文字数(唯一需要关心的结果) |
| integer | 文本总字符数(含标点、空格等全部字符) |
| number | 中文占全部字符的比例,保留 4 位小数 |
| object | 分项计数: |
| string | 失败原因;成功时为空字符串 |
调用 count_chinese_characters(text="你好,World 2026!") 的返回:
{
"chinese_count": 2,
"total_characters": 14,
"chinese_ratio": 0.1429,
"breakdown": {
"chinese": 2,
"letters": 5,
"digits": 4,
"punctuation": 2,
"spaces": 1,
"other": 0
},
"error": ""
}2. count_chinese_characters_batch
批量统计多条文本,并给出合计,适合一次校验多段文案。
参数 | 类型 | 必填 | 说明 |
| array | 是 | 文本列表,最多 500 条,每条上限 200000 个字符 |
返回字段:results(与输入顺序一致,每项含 index / total_characters /
chinese_count / chinese_ratio)、text_count、total_chinese_count、error。
3. extract_chinese_text
抽取文本中的中文字符,剔除标点、空格、英文、数字等其它内容。
参数 | 类型 | 必填 | 说明 |
| string | 是 | 待处理文本,长度上限 200000 个字符 |
调用 extract_chinese_text(text="Hello 世界 2026") 的返回:
{
"chinese_text": "世界",
"chinese_count": 2,
"error": ""
}三个工具都是只读、幂等操作(MCP readOnlyHint / idempotentHint 已标记为 true),
客户端可以安全地自动调用。
计数规则
计入中文:
CJK 统一表意文字基本区(U+4E00–U+9FFF)与扩展 A–I 区
兼容表意文字(U+F900–U+FAFF、U+2F800–U+2FA1F)
表意数字零"〇"(U+3007)
不计入:
中文标点与英文标点:
,。、;:!?""''()《》、,.;:!?()<>等空白:空格、制表符、换行
英文字母与其它拉丁字母、阿拉伯数字
emoji、各类符号
其它文字系统:日文假名、韩文谚文、西里尔字母等(归入
other/letters)
已知边界(刻意为之,不视为缺陷):
日文汉字与中文汉字同属 CJK 表意文字区,Unicode 层面无法区分,一律计入中文。
日文迭字符"々"(U+3005)不是表意文字本体,不计入。
全角数字"123"属于数字,不计入;汉字数字"一二三"计入。
项目结构
chinese-char-counter-mcp/
├── src/chinese_char_counter_mcp/
│ ├── counter.py # 计数核心:纯函数,无副作用,可单独复用
│ ├── server.py # MCP 工具定义与显式注册、控制台入口
│ └── __main__.py # python -m 入口
├── npm/ # 零依赖 Node 实现(npm 包,供 npx 使用)
│ ├── bin/server.js # MCP 服务入口(STDIO)
│ ├── lib/counter.js # 计数核心(与 Python 版逐字段对齐)
│ ├── lib/protocol.js# 极简 MCP JSON-RPC 服务端
│ └── test/ # node --test:计数规则 + 协议链路
├── tests/ # pytest 单元测试(计数规则 + 工具信封契约)
├── scripts/
│ └── e2e_stdio_client.py # STDIO 端到端验证:initialize -> list_tools -> call_tool
├── docs/publish-guide.md # 发布到 GitHub / PyPI / ModelScope MCP 广场的步骤
└── pyproject.toml开发
# Python 实现
pip install -e ".[dev]"
pytest -q # 单元测试
python scripts/e2e_stdio_client.py # STDIO 端到端验证(需已安装 mcp)
# Node 实现
cd npm && npm test # node --test:计数规则 + STDIO 协议全链路设计约定:
counter.py只做纯计算,不 importmcp,方便复用与测试。server.py的工具永远返回 dict 信封、永不向 MCP 层抛裸异常:成功时error为空字符串, 失败时error给出可直接阅读的原因,其余字段类型保持稳定。服务端不向 stdout 打印任何内容(STDIO 传输里 stdout 是协议通道),日志走 stderr。
在 ModelScope(魔搭)MCP 广场上架
本仓库已按魔搭"从 GitHub 仓库快速创建"的要求准备:根目录 README 正文中包含可解析的
STDIO 服务配置(即上面的 mcpServers JSON 块),command 为 uvx,包已发布到 PyPI。
完整步骤与部署检测自查清单见 docs/publish-guide.md。
License
Available Tools
3 toolscount_chinese_charactersARead-onlyIdempotent
统计一段文本的中文字数(不算标点、空格、英文字母、数字等)。
Args: text: 待统计的文本,长度上限 200000 个字符。
Returns: chinese_count 为中文字数(CJK 表意文字,含扩展区与“〇”); total_characters 为文本总字符数(含标点、空格等全部字符); chinese_ratio 为中文占全部字符的比例(保留 4 位小数); breakdown 为分项计数:chinese/letters/digits/punctuation/spaces/other; error 为失败原因,成功时为空字符串。
Inputs: text (required). Returns: error, breakdown, chinese_count, chinese_ratio, +1 more
count_chinese_characters_batchARead-onlyIdempotent
批量统计多条文本的中文字数,并给出合计。
Args: texts: 文本列表,最多 500 条,每条长度上限 200000 个字符。
Returns: results 与输入顺序一致,每项含 index/total_characters/chinese_count/chinese_ratio; text_count 为实际统计条数;total_chinese_count 为所有文本的中文字数合计; error 为失败原因,成功时为空字符串。
Inputs: texts (required). Returns: error, results, text_count, total_chinese_count
extract_chinese_textARead-onlyIdempotent
抽取文本中的中文字符,剔除标点、空格、英文、数字等其它字符。
Args: text: 待处理的文本,长度上限 200000 个字符。
Returns: chinese_text 为仅由中文字符按原顺序组成的文本; chinese_count 为抽取出的中文字数; error 为失败原因,成功时为空字符串。
Inputs: text (required). Returns: error, chinese_text, chinese_count
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 3 tools
The three tools have related but fairly distinguishable purposes: single counting, batch counting, and Chinese-only extraction. There is some overlap between count_chinese_characters and count_chinese_characters_batch, but the batch version is clearly scoped for multiple texts.
All tool names follow a consistent verb_noun-ish snake_case pattern: count_chinese_characters, count_chinese_characters_batch, extract_chinese_text. The relationship between the count tools is also clearly signposted by the _batch suffix.
Three tools is a reasonable, focused set for a Chinese character counter server. It is not thin for the core purpose, though batch counting could arguably be folded into the main count tool.
The server covers counting, batch counting, and extracting Chinese characters, which is solid for the domain. There are minor gaps around broader text-processing needs such as normalization or validation, but the stated purpose is well covered.
Maintenance
Related MCP Connectors
Calculate Tian Zi Ge grids, decompose Hanzi strokes, grade HSK 3.0 words & convert Pinyin tones.
61Count occurrences of any character in your text instantly. Specify the character and get precise c…
Exact character/word counting, reversal, palindrome checks, indexing, sorting; Unicode-safe.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables counting characters or bytes in text with options to include or exclude whitespace. Provides a simple tool for text analysis and length measurement.1MIT
- AlicenseNot gradedqualityDmaintenanceProvides tools for AI models to count characters and words in text, supporting English and other space-delimited languages.1MIT
- AlicenseNot gradedqualityDmaintenanceMCP server that counts the number of Chinese characters (excluding punctuation, spaces, English, and numbers) in a given text, compatible with Streamable HTTP and JSON-RPC.MIT
- FlicenseNot gradedqualityCmaintenanceAccurately counts characters, bytes, and manuscript paper based on Unicode grapheme clusters, and provides deterministic feedback to help AI meet exact length limits for self-introductions or school records.-