office-oxide-mcp-server

Use office_list_pdf_fields to retrieve exact field names from the PDF, then call office_fill_pdf_form with a fields object mapping those names to values. Field names are case-sensitive, so always list them first before filling.

Call the office_read tool with the file path and output_format set to json. The response returns structured sheets with cells containing values and formulas, which you can process or transform in downstream steps.

Yes, it runs as a stdio MCP server and can be registered in Claude Desktop, Cursor, and VS Code by adding the office-oxide-mcp command with the --transport stdio argument to the client's MCP configuration file.

The most common cause is mismatched field names, since they are case-sensitive. Run office_list_pdf_fields first to get the exact names, and for flat PDFs without form fields use office_analyze_pdf_layout plus office_overlay_pdf_text instead.

Very large spreadsheets can slow processing. Use the chunks output format to process incrementally, read only needed sheets, and consider splitting files with more than 10 million cells into multiple documents.