Markitdown Dingding

Get Cookie:

• Node.js version reads from dingtalk_cookies.json automatically
• Run npm run cookie:login in the nodejs directory for automatic browser login
• Or manually copy cookie from browser DevTools

2. Markitdown Installation

pip install markitdown

MCP Tools Available

After configuration, these tools will be available:

Workflow

Basic Flow

DingTalk URL/ID → MCP parse_document → HTML → markitdown → Markdown

Step-by-Step

1. Get DingTalk document URL or NODE_ID from user

   • Full URL: https://alidocs.dingtalk.com/i/nodes/xxxxx
   • Or just NODE_ID: R1zknDm0WR3e100Zc2on3deQVBQEx5rG

2. Call MCP tool to fetch document:

   # Option A: parse_document (saves files)
   result = mcp_dingtalk_doc_parse_document(
       url_or_node_id="https://alidocs.dingtalk.com/i/nodes/...",
       save_files=True,
       output_dir="/Users/xdf/Documents/dingtalk_docs"
   )
   
   # Option B: get_html (content only)
   result = mcp_dingtalk_doc_get_html(
       url_or_node_id="doc_node_id"
   )
   

3. Convert HTML to Markdown:

   # If parse_document saved HTML file
   markitdown "/path/to/output/document.html" -o output.md
   
   # Or if you have HTML content directly
   echo "$html_content" > /tmp/doc.html
   markitdown /tmp/doc.html -o output.md
   

Usage Examples

Example 1: Full Document Extraction

# 1. Parse DingTalk document
result = mcp_dingtalk_doc_parse_document(
    url_or_node_id="https://alidocs.dingtalk.com/i/nodes/abc123",
    save_files=True
)

# 2. Convert HTML to Markdown
html_path = result['output_dir'] + '/document.html'
terminal(f"markitdown '{html_path}' -o /tmp/output.md")

# 3. Read markdown content
content = read_file("/tmp/output.md")

Example 2: Quick Content Extraction

# 1. Get HTML content directly
result = mcp_dingtalk_doc_get_html(
    url_or_node_id="document_node_id"
)

# 2. Save HTML to temp file
html_content = result['html']
write_file("/tmp/ding_doc.html", html_content)

# 3. Convert to Markdown
terminal("markitdown /tmp/ding_doc.html -o /tmp/ding_doc.md")

# 4. Return markdown
content = read_file("/tmp/ding_doc.md")

Error Handling

Common Issues

MCP Connection Issues

If MCP tools not appearing:

1. Check server is running: node /Users/xdf/Documents/XDF/mcp-dingtalk-doc/nodejs/dist/index.js
2. Verify config syntax in ~/.hermes/config.yaml
3. Restart Hermes Agent after config changes

Best Practices

1. Use parse_document when you need:

   • Complete document with images
   • Preserved formatting
   • Local file archive

2. Use get_html when you need:

   • Quick content extraction
   • No file saving
   • In-memory processing

3. Always check output:

   • Review markdown for formatting issues
   • Verify images are properly referenced
   • Check for DingTalk-specific formatting

References

• DingTalk MCP Server
• Markitdown GitHub