Schema-aware MCP server for Redfish APIs
This server demonstrates basic MCP tool registration with 3 read-only tools targeting Redfish infrastructure APIs. However, several critical gaps prevent a higher score: (1) All tool descriptions are extremely terse (10-40 chars), well below the 50-200 char LLM-optimized baseline and below the 20-char minimum threshold for maintaining score above 0. (2) Two of three tools have empty input schemas ({}), providing no guidance on optional parameters or constraints. (3) No structured output schemas are documented, the test script validates responses empirically, but the tool definitions themselves declare no return types. (4) Parameter descriptions are absent for the one tool (list_systems) that has a parameter. (5) No error recovery guidance is present, errors would surface to the LLM with no next-step hints. The naming is acceptable ('list_', 'service_info', consistent verbs), and the tools are read-only with clear risk classification, which accounts for the modest score. The server appears functionally correct (validation script passes) but is poorly documented for agent reasoning.
List available Redfish endpoints
List compute systems managed by Redfish service
Get service root information and capabilities
Tool descriptions are extremely terse (10 - 40 characters), far below the 50 - 200 character LLM-optimized baseline. LLMs cannot reliably reason about tool selection with descriptions this brief.
Two of three tools (list_endpoints, service_info) declare empty input schemas ({}), providing no indication of optional parameters, defaults, or constraints. Agents have no guidance on what inputs to provide.
No output/return schemas are documented in tool definitions. Agents cannot plan downstream tool calls or data extraction without knowing what fields the responses contain. The validation script checks responses empirically, but the MCP tool metadata itself declares no return types.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | F | 39 | 2026-07-28+ | v2 |
The list_systems tool has a parameter 'include_details' with no description. LLMs cannot infer what effect this boolean toggle has on the response.
No error recovery guidance is documented. If a tool call fails, the LLM receives no hint about whether to retry, ask the user, or try an alternative tool.
No pagination or limit parameters are visible on list-returning tools (list_endpoints, list_systems). The validation script checks 'len(endpoints) > 0' and systems have 'truncated' flags, suggesting partial results may occur, but the tools provide no way to control page size or offset.