cnki-export

name	cnki-export
description	Export paper from CNKI and push to Zotero, or save as RIS file. Use when user wants to save a paper to Zotero or export citation data.
argument-hint	[zotero|ris|gb] [paper URL or blank if on detail page]
version	0.1.0

CNKI Export & Zotero Integration

Export paper citation data from CNKI and push directly to Zotero, or save as RIS file.

Arguments

• zotero (default) — push to Zotero desktop via local API
• ris — save as .ris file
• gb — output GB/T 7714 citation text
• Optionally include a paper URL

Mode Selection

Choose the right mode based on context:

Context	Mode	Tool calls
On a paper detail page	Single export (Step 1A)	1 evaluate + 1 bash = 2
On a search results page, save all/selected	Batch export (Step 1B)	1 evaluate + 1 bash = 2
Need to search then save	Use cnki-search first, then batch export	4 total

Always prefer batch export (1B) when multiple papers need saving. It avoids navigating to each detail page (saves ~3 calls per paper).

Steps

1A. Single export: from paper detail page

Use mcp__chrome-devtools__evaluate_script:

async () => {
  const url = document.querySelector('#export-url')?.value;
  const params = document.querySelector('#export-id')?.value;
  const uniplatform = new URLSearchParams(window.location.search).get('uniplatform') || 'NZKPT';
  if (!url || !params) return { error: 'Not on a paper detail page' };

  const resp = await fetch(url, {
    method: 'POST',
    headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
    body: new URLSearchParams({ filename: params, displaymode: 'GBTREFER,elearning,EndNote', uniplatform })
  });
  const data = await resp.json();
  if (data.code !== 1) return { error: data.msg };

  const result = {};
  for (const item of data.data) {
    result[item.mode] = item.value[0];
  }

  const body = document.body.innerText;
  result.pageUrl = window.location.href;
  result.issn = body.match(/ISSN[：:]\s*(\S+)/)?.[1] || '';
  result.dbcode = document.querySelector('#paramdbcode')?.value || '';
  result.dbname = document.querySelector('#paramdbname')?.value || '';
  result.filename = document.querySelector('#paramfilename')?.value || '';

  return result;
}

1B. Batch export: from search results page (PREFERRED for multiple papers)

On any CNKI search results page, extract checkbox values and call the export API directly — no need to navigate to detail pages.

Key discovery: input.cbItem checkbox value === detail page #export-id (same encrypted ID).

Use mcp__chrome-devtools__evaluate_script:

async () => {
  const API_URL = 'https://kns.cnki.net/dm8/API/GetExport';

  // Get all checkbox values (= export encrypted IDs)
  const checkboxes = document.querySelectorAll('.result-table-list tbody input.cbItem');
  const rows = document.querySelectorAll('.result-table-list tbody tr');

  if (checkboxes.length === 0) return { error: 'No results on page' };

  const allPapers = [];
  for (let i = 0; i < checkboxes.length; i++) {
    const exportId = checkboxes[i].value;
    const paperUrl = rows[i]?.querySelector('td.name a.fz14')?.href || '';

    const resp = await fetch(API_URL, {
      method: 'POST',
      headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
      body: new URLSearchParams({ filename: exportId, displaymode: 'GBTREFER,elearning,EndNote', uniplatform: 'NZKPT' })
    });
    const data = await resp.json();
    if (data.code === 1) {
      const result = {};
      for (const item of data.data) { result[item.mode] = item.value[0]; }
      result.pageUrl = paperUrl;
      // Extract ISSN from ENDNOTE %@ field
      const issnMatch = result.ENDNOTE?.match(/%@\s*([^\s<]+)/);
      result.issn = issnMatch ? issnMatch[1] : '';
      result.dbcode = 'CJFQ';
      result.dbname = '';
      result.filename = '';
      allPapers.push(result);
    }
  }

  return allPapers; // JSON array, directly writable to file for Python script
}

To export only specific papers (e.g. #1, #3, #5), filter by index:

// Replace the for loop condition:
const indices = [0, 2, 4]; // 0-indexed: papers #1, #3, #5
for (let i = 0; i < checkboxes.length; i++) {
  if (!indices.includes(i)) continue;
  // ... rest same
}

2. Push to Zotero

Save the export data (single object or JSON array) to a temp file, then run the Python script:

python scripts/push_to_zotero.py /tmp/papers.json

The Python script handles both single paper {} and batch [{}, {}, ...] JSON input.

• UTF-8 encoding (avoids Windows encoding issues)
• Parsing ELEARNING format into Zotero item fields
• Calling POST http://127.0.0.1:23119/connector/saveItems
• Returns: 201 = success, 500 = error, 0 = Zotero not running

3. Report result

Single:

已将论文添加到 Zotero:
  标题: {title}
  作者: {authors}
  期刊: {journal}

GB/T 7714 引用: {gbt_citation}

Batch:

已批量添加 {count} 篇论文到 Zotero:
  1. {title1} ({journal1})
  2. {title2} ({journal2})
  ...

Export API Reference

Parameter	Value	Source
API URL	https://kns.cnki.net/dm8/API/GetExport	Fixed, works from any page
filename	Encrypted ID	Detail page: #export-id; Results page: input.cbItem value
displaymode	GBTREFER,elearning,EndNote	Comma-separated modes
uniplatform	NZKPT	Required

Verified selectors

Element	Selector	Page
Export URL	#export-url	Detail page only
Export ID	#export-id	Detail page only
Checkbox (= export ID)	input.cbItem	Search results page
Result rows	.result-table-list tbody tr	Search results page
Title link	td.name a.fz14	Search results page

Zotero API Reference

POST http://127.0.0.1:23119/connector/saveItems
Content-Type: application/json
X-Zotero-Connector-API-Version: 3

Response: 201 = created, 500 = error Collection: Saves to Zotero's currently selected collection.

Query collections:

python scripts/push_to_zotero.py --list

Important Notes

• Windows encoding: Must use Python script, cannot pass Chinese JSON via bash/curl directly
• Zotero must be running: localhost:23119 requires Zotero desktop in background
• Chinese authors: Use name field (single field, not split), creatorType: "author"
• Batch export saves ~90% tool calls: 9 papers: 33 calls → 3 calls
• CNKI Export API: filename must be encrypted ID (#export-id or input.cbItem value), NOT #paramfilename