Loading ToolBench
Example MCP server with tools demonstrating phrases about animals
This server has severe definition quality issues that make it unsuitable for production use. Both tools have identical, misleading descriptions ('Prints a phrase about a cat' for both about_mouse and about_cat), zero input schema complexity (empty objects), and no parameter documentation. The naming is vague (about_mouse, about_cat do not start with action verbs like 'get_', 'describe_', or 'retrieve_'), and the server demonstrates minimal effort in tool design patterns. These are toy tools with no real-world applicability.
Prints a phrase about a cat.
Prints a phrase about a cat.
Tool 'about_mouse' has description 'Prints a phrase about a cat.', this is factually incorrect (describes a mouse, not a cat) and is identical to about_cat description. Misleading descriptions cause LLMs to select the wrong tool. [Baseline: 100% of A+ tools have clear, accurate descriptions]
Both tools lack action verbs in names. 'about_mouse' and 'about_cat' do not start with verbs (get_, describe_, retrieve_, list_). Action verbs let LLMs infer intent from the name alone. Baseline: 90% of A+ tools start with action verbs. These fail to match production naming conventions.
Both tools have empty input schemas (Input: {}). No parameters are defined, documented, or constrained. While these tools take no parameters, the schema is not explicitly documented in the observable code as a formal JSON Schema. Schemas MUST be visible and properly typed.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 0 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 15 | - | v1 |
No output schema is documented. LLMs cannot determine what fields to expect. The return type is a string ('O rato roeu a roupa do rei da china!'), but this is not formally declared in a schema. [Baseline: 100% of A+ tools have documented return types]
No error handling. If the function raises an exception (syntax error, memory issue, etc.), there is no try-catch, no error categorization (retryable vs. fatal), and no recovery guidance. LLMs receive raw exceptions with no actionable next steps.
Trivial tool purpose. Both tools merely print hard-coded Portuguese phrases about animals. They have no real-world use case, no dynamic parameters, no external data dependencies, and no composition with other tools. This appears to be an example/template, not a functional server.