Multi-server MCP implementation for Smart Grid transformer asset operations, diagnostics, and maintenance. Comprises four specialized servers: IoT (sensor telemetry), FMSR (failure mode diagnosis), TSFM (time-series forecasting), and WO (work order management).
19 tools with consistent verb-noun naming (list_, get_, create_, update_, search_, detect_, forecast_, trend_, estimate_). All tools have descriptions (avg ~80 chars, within baseline 34-392). Input schemas present for all tools with typed parameters. However, descriptions are terse and lack LLM-optimized guidance on WHEN to use each tool vs. alternatives. Output schemas are not documented, LLMs cannot plan downstream calls. Error handling is absent from descriptions. No tool annotations (readOnlyHint/destructiveHint). Parameter descriptions are minimal; many lack format/range guidance. No pagination documented for list tools despite potential large result sets.
Classify a set of gas concentrations into a fault type using the IEC 60599 Rogers Ratio method.
Create a new maintenance work order for a transformer.
Detect anomalous sensor readings using a rolling z-score method.
Estimate repair downtime based on fault type and severity.
Project the RUL forward by `horizon_days` using a linear degradation model.
Return full nameplate and status metadata for a single transformer.
Retrieve the most recent dissolved gas analysis (DGA) record for a transformer.
Output schemas not documented. LLMs cannot infer what fields are returned or plan downstream tool calls. E.g., does list_assets return asset_id, name, status, health_score? Does get_sensor_readings return timestamps, values, units?
Descriptions lack LLM-optimized guidance. Most are 50-80 chars and state WHAT the tool does but not WHEN to use it or how it differs from similar tools. E.g., list_assets vs. get_asset_metadata distinction is unclear. No dependency hints (e.g., 'Call list_sensors() first to discover valid sensor IDs').
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 61 | 2026-07-28+ | v2 |
Retrieve a single historical fault record by its ID.
Return the most recent remaining useful life (RUL) estimate for a transformer.
Return the sensors and gases most strongly associated with a failure mode.
Return time-series readings for one sensor on one transformer.
List all Smart Grid transformer assets.
Return the full catalogue of known transformer failure modes.
Browse historical fault records from the Smart Grid dataset.
List all sensor IDs available for a given transformer.
List work orders created by the agent in this session.
Search failure modes by keyword (name, description, gas, or IEC code).
Compute the trend (slope) of a sensor's readings over a time window.
Update an existing work order's status, priority, assignee, or add a note.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). LLMs cannot distinguish safe read-only tools from destructive ones. create_work_order and update_work_order should be marked destructiveHint=true; all others readOnlyHint=true.
Pagination not documented for list tools. list_assets, list_sensors, list_failure_modes, list_fault_records, list_work_orders lack limit/offset or cursor parameters in visible schema. If results can exceed 100 items, LLMs cannot fetch all data without explicit pagination guidance.
No error handling guidance in descriptions. Tools do not document what errors can occur, whether they are retryable, or what the LLM should do next. E.g., create_work_order could fail if transformer_id is invalid, should the LLM call list_assets first?