Loading ToolBench
MCP server for AI data center design and sizing via the AIDC-AI.IO API
Single tool 'design' has a reasonable parameter schema with types and descriptions, but the tool definition itself lacks crucial context. The tool description is adequate (137 chars, within 10-1024 range) and explains the core function. However, the output schema is NOT documented anywhere in the source, the example shows sample output fields but no formal schema definition for what the tool returns. This is a critical gap for production use. Parameter types are present (number, string, object) and descriptions exist for top-level params, but the nested 'options' object lacks detailed constraints (no enum values for redundancy/coolingMode, no explicit range for pueTarget). No error handling guidance is visible. The tool does only one thing (design a data center), which is good composition, but the lack of output schema documentation and missing parameter constraints significantly reduce confidence.
Sizes an AI data center given IT load, rack density, GPU generation, site area, region, and design options (redundancy, cooling mode, PUE target). Returns rack count, PUE design, total MVA, cooling load distribution, CDU count, estimated cost, and build timeline.
Output schema not documented. The tool description mentions it 'Returns rack count, PUE design, total MVA, cooling load distribution, CDU count, estimated cost, and build timeline' as free text, but no formal JSON Schema is provided for the return type. LLMs cannot reliably extract and chain these fields without a documented schema.
Nested parameter 'options' lacks constraints. The object properties (redundancy, coolingMode, pueTarget) have descriptions but no enum constraints for redundancy ('n_plus_1' is the only example, are other values valid?) and coolingMode ('liquid' shown, is 'air' also valid?). pueTarget lacks min/max bounds.
No error handling guidance visible. If a call fails (e.g. invalid region, site area too small, incoherent redundancy option), the source provides no indication of what errors to expect, how to interpret them, or recovery steps. LLMs cannot self-correct without this information.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 50 | 2026-07-28+ | v2 |
Parameter naming ambiguity: 'siteAreaSqm' is clear, but 'region' as a free-form string invites hallucination. The example shows 'metropolitan', but are other values like 'rural', 'coastal', 'temperate' valid? No enum constraint provided.
Tool name 'design' is generic and verb-less. While the context (AIDC data center) clarifies intent, a more specific name like 'design_data_center' or 'size_data_center' would be self-documenting and less ambiguous if this MCP server is integrated alongside other tools.