cnki-search
<p align="center">
<img src="docs/assets/logo.png" width="160" alt="cnki-search" />
</p>
<h1 align="center">cnki-search</h1>
<p align="center">
从<a href="https://search.cnki.com.cn/">知网空间</a>检索中文文献题录,通过 MCP 提供给各类 Agent。<br/>
返回题名、作者、年份、文献类型和详情页链接;不提供全文下载。
</p>
<p align="center">
<a href="https://modelcontextprotocol.io/"><img src="https://img.shields.io/badge/MCP-stdio-555?labelColor=111" alt="MCP" /></a>
<a href="https://github.com/deepseek-ai/deepseek-harness"><img src="https://img.shields.io/badge/DSH-plugin-1a73e8?labelColor=111" alt="DSH plugin" /></a>
<a href="https://www.python.org/"><img src="https://img.shields.io/badge/python-3.12%2B-3776ab?labelColor=111" alt="Python 3.12+" /></a>
<a href="LICENSE"><img src="https://img.shields.io/badge/license-MIT-2ea44f?labelColor=111" alt="MIT" /></a>
</p>
数据来自 `search.cnki.com.cn` 的公开检索接口,无需知网账号。检索范围是知网空间,不是 KNS 全部馆藏。
## 功能
| 工具 | 作用 |
|---|---|
| `search` | 按主题、篇名、关键词、全文或摘要检索;可限定作者、导师、年份、文献类型 |
| `get_record` | 根据检索得到的知网 / CDMD 链接读取摘要与学位信息 |
`search` 参数:
| 参数 | 取值 | 说明 |
|---|---|---|
| `query` | 字符串 | 检索词。仅查作者或导师时可为空 |
| `field` | `theme` / `title` / `keyword` / `content` / `summary` | 主题 / 篇名 / 关键词 / 全文 / 摘要 |
| `page` | 1–50 | 页码 |
| `author` | 字符串 | 作者 |
| `advisor` | 字符串 | 导师 |
| `year` | 如 `2023` | 年份 |
| `kind` | `all` / `journal` / `thesis` / `phd` / `master` | 全部 / 期刊 / 博硕 / 博士 / 硕士 |
| `sort` | `relevance` / `date` / `downloads` / `cites` | 相关度 / 发表时间 / 下载 / 被引 |
已知题名时用 `field=title`。`keyword` 按词拆分,容易混入不相关结果。
返回 `status`:`ok` 为正常;`blocked` 为接口拒绝或校验页;`error` 为参数或网络错误。不要把 `blocked` 当成零结果。
## 安装
请先让 Agent 阅读本仓库并按其中说明完成安装:
https://github.com/Biogod2020/cnki-search
需要本机已安装 [uv](https://github.com/astral-sh/uv)(Python 3.12+)。Agent 读完后可用下面命令启动,无需事先 clone:
```bash
uvx --from git+https://github.com/Biogod2020/cnki-search.git cnki-search
```
进程在 stdio 上等待 MCP 宿主,没有交互提示。
开发与测试:
```bash
git clone https://github.com/Biogod2020/cnki-search.git
cd cnki-search
uv sync --extra dev
uv run pytest -m "not live"
```
联网回归:`RUN_LIVE_CNKI=1 uv run pytest -m live -s`
## 接入 MCP
把仓库链接交给 Agent,或把下面配置写入宿主。宿主会通过 `uvx` 拉取并运行。
```json
{
"mcpServers": {
"cnki": {
"command": "uvx",
"args": [
"--from",
"git+https://github.com/Biogod2020/cnki-search.git",
"cnki-search"
]
}
}
}
```
适用于 Claude、Cursor、Codex 及其他 MCP stdio 宿主。工具名:`search`、`get_record`。
## 接入 DeepSeek Harness
与 [dsh-bing-search](https://github.com/Biogod2020/dsh-bing-search) 相同:由 `@deepseek-ai/dsh-mcp-client` 拉起本仓库的 stdio 服务。发现后的工具名为:
```text
mcp__cnki__search
mcp__cnki__get_record
```
写入 `cordis.yml` 可用 [`examples/dsh.cordis.yml`](examples/dsh.cordis.yml)。写入 `$DSH_HOME/profiles/<name>/cordis.patch.yml` 时必须用 `insert`,否则 id 不存在会被静默跳过。见 [`examples/dsh.patch.yml`](examples/dsh.patch.yml):
```yaml
- insert:
- id: mcp-cnki
name: '@deepseek-ai/dsh-mcp-client'
config:
serverName: cnki
transport: stdio
command: uvx
args:
- --from
- git+https://github.com/Biogod2020/cnki-search.git
- cnki-search
toolCallTimeoutMs: 30000
failOnStartupError: true
reconnect:
enabled: true
initialDelayMs: 500
maxDelayMs: 30000
maxAttempts: 10
```
## 调用示例
```json
{
"query": "SCA3发病年龄的临床预测模型构建与罕见变异关联研究",
"field": "title"
}
```
```json
{
"status": "ok",
"provider": "cnki",
"returned_count": 1,
"records": [
{
"title": "SCA3发病年龄的临床预测模型构建与罕见变异关联研究",
"url": "https://cdmd.cnki.com.cn/Article/CDMD-10533-1025564694.htm",
"authors": "彭林柳",
"year": "2023",
"kind": "博士论文"
}
]
}
```
```json
{
"query": "空间转录组",
"field": "title",
"kind": "thesis",
"sort": "date"
}
```
Python:
```python
from cnki_scholar import search_cnki, get_record
hits = search_cnki("影像组学 阿尔茨海默病", field="title", kind="thesis")
detail = get_record(hits.records[0].url)
```
## 说明
- 覆盖范围为知网空间,新入库学位论文可能检索不到。
- 期刊可用接口的文献类型参数;博硕类型在解析后过滤。
- 列表页结构若变更,解析会失效。
- 文献版权归知网及原作者。请遵守知网使用条款与所在机构规定。
## 许可
MIT
TDQS
Scored across 2 tools
The two tools are distinctly separated: 'search' performs queries across metadata fields, while 'get_record' fetches details from a specific URL returned by search. There is no functional overlap between them.
Both tool names follow a clear lowercase_with_underscores convention and use verb-first naming ('search', 'get_record'). The pattern is consistent and predictable.
With only two tools, the server is minimal. The scope (searching and retrieving CNKI records) is narrow enough to justify this count, but it borders on being too thin for a productive search API.
The server covers the essential metadata search and record retrieval workflows. It lacks features like citation export or full-text download, but for its stated purpose of metadata access, the coverage is adequate with only minor gaps.