office-oxide-mcp-rust-document-processor

Call office_create_xlsx to make the workbook, office_write_range to insert data, office_format_range for number formatting, and office_create_chart with a data range and chart type like column, bar, line, or pie. Batch writes with office_write_range are faster than individual cell writes.

First call office_list_pdf_fields to get exact case-sensitive field names, then call office_fill_pdf_form with a fields object mapping names to values and an output path. For flat PDFs without form fields, use office_analyze_pdf_layout and office_overlay_pdf_text to place text at coordinates.

The server works with Claude Desktop, Cursor, and VS Code Copilot via stdio transport. Add it to the client's MCP configuration with the command office-oxide-mcp and args --transport stdio, using the full binary path if it is not in PATH.

Yes, the office_read tool accepts an output_format parameter of json, markdown, text, or chunks. JSON returns structured sheet data as rows of key-value objects, while chunks mode suits RAG embedding pipelines and large files.

Field names are case-sensitive and must match exactly what office_list_pdf_fields returns. If the PDF has no AcroForm or XFA fields, it is a flat document and requires office_overlay_pdf_text with page coordinates instead.

Use office_read with output_format set to chunks and a chunk_size such as 1000 rows to process data incrementally. You can also increase the stack size with ulimit -s 16384 and set OFFICE_OXIDE_MAX_CONCURRENT to control concurrency.