Loading ToolBench
A FastMCP server that exposes tools for reading and exploring Excel (.xlsx) files.
The Excel MCP Server provides three functional read-only tools with clear, complete naming and documented parameter schemas. All three tools have descriptions and input schemas with type information. However, there are gaps in output schema documentation, parameter descriptions lack formatting constraints, and error handling does not guide recovery actions. The tools follow basic composition patterns but lack some LLM-optimization details.
Show the dimensions (rows × columns) of every sheet in an Excel workbook.
List every sheet name present in an Excel workbook.
Fetch rows from one sheet or from all sheets in an Excel workbook. The tool auto-detects the number of columns in each sheet so you never need to specify column indices manually. The first row of every sheet is treated as the column header row; subsequent rows become data records.
Output schemas not formally documented in tool definitions. Docstrings describe return structure, but JSON Schema exposed via MCP protocol does not include 'returns' or outputSchema fields. LLMs cannot reliably infer the output shape from signatures alone.
Parameter descriptions lack formatting constraints. For example, file_path accepts 'string|null' but does not specify: file must be .xlsx, path must be absolute or relative, max path length, etc. n_rows defaults to 10 but has no documented bounds (min, max).
Error handling provides no recovery guidance. Errors like FileNotFoundError or ValueError are raised with basic messages, but do not guide the LLM on next steps.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 57 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 45 | - | v1 |
No pagination support. If an Excel sheet has 10,000 rows and n_rows=10, there is no way to fetch rows 11 - 20.