Loading ToolBench
A Model Context Protocol server implementing tools for basic operations, resources for server metadata, and prompts for summarization tasks.
V.A.L MCP server provides two minimalist tools with adequate but basic definitions. Both tools have descriptions and input schemas present, but descriptions are very short (10-34 chars), parameter descriptions are minimal or absent, and there is no output schema documentation. The 'ping' tool lacks any input parameters and serves only as a heartbeat; the 'add' tool has properly typed float parameters with brief descriptions but no guidance on valid ranges, constraints, or output structure. No error handling guidance, no security considerations, and no composition across tools. The server includes prompts and resources (summarize prompt, server_info resource) which are present but not thoroughly evaluated as tools. Overall, definitions meet the bare minimum for tool registration but fall short of production quality.
Add two numbers and return their sum.
Return a heartbeat response from the server.
Tool descriptions are extremely brief (10-34 characters), well below the rubric baseline of 194 chars (avg) and minimum 50 chars for actionable context. 'ping' and 'add' descriptions lack guidance on when to call, prerequisites, or output structure.
No output schema documented for any tool. LLMs cannot plan downstream calls or extract return values without knowing the shape of responses. The 'ping' tool returns a string; the 'add' tool returns a float, but this is implicit, not explicit in the schema.
Parameter descriptions for 'add' are minimal ('First number', 'Second number') and lack constraint information. No mention of valid ranges, precision expectations, or whether negative numbers are allowed. Rubric requires format, range, and allowed values in parameter descriptions.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 44 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 34 | - | v1 |
No error handling guidance. Neither tool documents what happens on invalid input (e.g., non-numeric input to 'add', or what errors should be retried vs. are fatal). Rubric requires error responses to tell LLMs what to do next and categorize errors as retryable, user-fixable, or fatal.
'ping' tool lacks semantic clarity. Named 'ping' (not 'check_server_health' or 'get_server_status'), and description does not explain when or why to call it instead of querying the 'server://info' resource. Ambiguous naming and positioning relative to the resource.