Loading ToolBench
MCP server for oee: Overall Equipment Effectiveness, the six big losses and charts for AI agents.
Server provides 12 tools with complete JSON Schema definitions and descriptions. Naming follows verb_noun patterns (compute_oee, aggregate_oee, reliability). Descriptions are present but brief (avg 60 chars, below 194-char baseline). All tools are READ_ONLY with no state-modifying operations. Input schemas are well-structured with Pydantic models and proper type hints. However, descriptions lack actionable context (WHEN to use, prerequisites, dependencies). Error handling returns generic hints rather than recovery guidance. Output schemas are documented via to_dict() + summary() but not explicitly declared in tool definitions.
Roll OEE up across machines or shifts correctly (sums the buckets).
Takt time and the required rate; with a cycle time, whether it keeps up.
OEE, the time waterfall and the six big losses from times and counts.
The input fields, units and OEE definitions, to format input correctly.
Render a Pareto of the six big losses as a PNG image.
The availability, performance and quality losses as lost units and money.
OEE from the three factors directly (each between 0 and 1).
Descriptions are too brief (avg 50 chars vs 194-char baseline). Missing WHEN to use, prerequisites, and dependencies. E.g., 'OEE from an event log' does not explain when to call this vs compute_oee or oee_from_factors.
Error handling returns generic hints ('call describe_inputs for the fields and units') instead of actionable recovery guidance. Does not categorize errors as retryable, user-fixable, or fatal.
Output schemas not explicitly documented in tool definitions. Callers must infer structure from to_dict() + summary() pattern. No declared return type schema in tool registration.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 55 | 2026-07-28+ | v2 |
OEE from an event log of production runs and downtime events.
MTBF, MTTR and inherent availability (the maintenance drivers of OEE availability). Give repair_times, or failures and total_repair_time.
Rolled throughput yield across steps (the product of the step yields).
Render the OEE and A/P/Q trend over a sequence of results as a PNG.
Render the OEE time waterfall (planned to fully productive) as a PNG.
Chart tools (waterfall_chart, loss_pareto_chart, trend_chart) return PNG bytes but no explicit documentation of output format, MIME type, or how to handle binary data in MCP responses.
No parameter validation constraints documented (e.g., numeric ranges, enum values for optional fields). LLMs may pass invalid values without clear feedback.