Loading ToolBench
A command-line tool for creating and deploying MCP servers
This server presents 10 tools across calculator, calendar, search, and weather domains. Tool definitions are present with reasonable descriptions (avg ~80 chars, within baseline range of 34-392). However, there are critical gaps: parameter descriptions are minimal or absent for many tools, input schemas lack proper constraint declarations (no enums for operation types, no range limits for numeric params), and output schemas are completely undocumented. Error handling and recovery guidance are absent. The server implements basic structure but falls short of production-grade quality due to missing schema rigor and underdocumented parameters. Tool naming is acceptable (verb_noun pattern) but some descriptions are too terse to guide LLM selection effectively.
Perform advanced mathematical operations.
Perform basic mathematical operations.
Create a new calendar event.
Delete a calendar event.
Get weather forecast for a location.
Search for images.
List events within a date range.
Missing output schema documentation. No tool specifies what fields are returned, their types, or structure. LLMs cannot plan downstream operations or extract needed data without knowing response schema.
Operation parameters use free-form strings instead of enums. 'operation' in basic_math accepts 'add', 'subtract', 'multiply', 'divide' but lacks enum constraint, LLM may hallucinate 'modulo' or 'power'. Same for advanced_math operations (sqrt, sin, cos, tan, log).
Numeric parameters lack bounds. 'numbers' array in basic_math and statistics has no min/max length. 'max_results' in search tools defaults to 5 but no constraint shown, unbounded on upper end. Latitude/longitude in get_forecast lack range validation (-90 to 90 for lat, -180 to 180 for lon).
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 47 | <=2025-11-25 | v2 |
| 2026-03-09 | F | 39 | - | v1 |
Search for news articles.
Calculate basic statistics.
Search the web for information.
Parameter descriptions are present but minimal. Most parameter descriptions are 5-15 words ('List of numbers to operate on', 'Duration in minutes') without format guidance, constraints, or context on when/why to provide values.
No error handling or recovery guidance. No documentation of what errors tools may return (e.g., division by zero in basic_math, invalid date formats in calendar tools, network failure in search/weather tools) or what the LLM should do next.
delete_event tool has no confirmation or dry-run step. Irreversible operations should support confirmation to prevent catastrophic agent mistakes (e.g., deleting all events in a calendar).
Tool descriptions lack context on WHEN to use each tool vs. similar tools. For example, web_search, news_search, and image_search are described identically ('Search the web/for news/for images') with no guidance on which to select for different query intents.
No pagination or result limits documented. list_events has no limit/offset/cursor parameters visible; search tools have 'max_results' default of 5 but no explicit cap or next_cursor support documented. Large result sets could blow context windows.
ISO format descriptions are informal. 'Start time in ISO format (YYYY-MM-DD HH:MM)' uses lowercase YYYY but real ISO 8601 is uppercase. List_events says 'ISO format (YYYY-MM-DD)' without clarifying timezone handling or whether times are UTC.