@negokaz/excel-mcp-server

Reads MS Excel files (.xlsx, .xls) and exposes sheet metadata (names, dimensions) plus cell-level data extraction via MCP protocol. Uses a Node.js Excel library (likely exceljs or xlsx) to parse binary/XML formats into in-memory workbook objects, then marshals cell values, formulas, and formatting into JSON-serializable structures for transmission over MCP transport. Supports multiple sheets within a single workbook with independent read operations per sheet.

Writes data to MS Excel files by accepting cell updates (value, formula, formatting) and sheet creation requests via MCP protocol. Loads existing workbooks into memory, applies mutations (cell writes, new sheets), and persists changes back to disk using the same underlying Excel library. Supports both appending to existing sheets and creating new sheets with initial data, with atomic write semantics per MCP call.

Implements MCP server specification to expose Excel read/write operations as callable tools with JSON schema definitions. Handles MCP message framing (stdio or HTTP transport), tool discovery, argument validation against schemas, and response serialization. Registers each Excel operation (read sheet, write cell, create sheet) as a distinct tool with typed parameters, enabling MCP clients (like Claude Desktop or custom agents) to discover and invoke Excel operations with IDE-like autocomplete and type checking.

Queries workbook structure to list all sheets with metadata (name, row count, column count, used range). Parses Excel file structure to extract sheet definitions without loading full cell data, enabling fast discovery of available sheets. Returns structured metadata that allows agents to understand workbook layout before performing targeted read operations, reducing unnecessary data transfer and improving query efficiency.

Extracts data from contiguous or non-contiguous cell ranges using A1 notation (e.g., 'A1:C10', 'A1,C1:C5') or row/column index tuples. Parses range specifications into cell coordinates, retrieves values from workbook, and returns as 2D arrays or object arrays with column headers. Supports both dense and sparse range queries, with optional header row interpretation for converting rows into key-value objects.