Scored across 156 tools
The set contains many overlapping families: multiple RCA/downtime tools (downtime_triage, downtime_root_cause, downtime_root_cause_live, downtime_attribution), several alarm-analysis tools (alarm_bad_actors, alarm_flood_analysis, alarm_event_clusters, alarm_cascade, alarm_rationalization_worksheet), many PLC program tools, and several baseline/OEE/data-quality variants. Descriptions often explain the differences, but an agent still faces a large ambiguous surface of near-neighbors and deprecated aliases.
Most names use snake_case and many protocol tools follow readable prefixes such as opcua_, modbus_, s7_, eip_, fins_, and ethercat_. However, the naming convention is not consistently verb_noun across the whole set, and there are confusing pairs such as health_summary vs opcua_health_summary and anomaly_scan vs opcua_anomaly_scan, plus deprecated aliases.
156 tools is far beyond an appropriate, well-scoped surface for one server. The count is inflated by protocol-specific readers, many analytics variants, deprecated aliases, and multiple overlapping RCA/alarm/PLC families, making the set difficult for an agent to navigate.
The toolkit covers a very broad industrial scope: multi-protocol reads, high-risk writes, historian access, OEE, RCA, alarm analytics, PLC program inspection, compliance, and UNS governance. There are still notable gaps, including no obvious OPC-UA or Modbus write operations and no full endpoint/tag configuration lifecycle through MCP, which leaves some workflows dependent on CLI or external tooling.
