MCP server for querying and writing data to InfluxDB, with support for time-series analysis, anomaly detection, and correlation analysis
Server has 8 tools with complete input schemas and descriptions. Naming follows verb_noun convention (list_, get_, write_, find_, query_, correlate). Descriptions are present (avg ~120 chars) but lack actionable context for LLM selection. Parameters have types and descriptions, but lack constraints (enums, ranges, patterns). No output schemas documented. Error handling is minimal, no recovery guidance or categorization. Security: write_value tool accepts user input without validation; flux injection mitigated via flux_escape() but not documented in tool description. Missing: pagination for list_measurements, field constraints for entity_id validation, per-tool error examples.
Pearson correlation coefficient between two sensors over a period -- both series are aligned to a common time grid via averaging before comparison.
Find anomalous sensor values (deviation from the mean by more than N sigma)
History of sensor values over a period. Returns the last N points from the range.
Latest value of a sensor by entity_id
List of fields for a specific measurement, with entity_id inside it
List of all measurements in the bucket, with entity_id inside each one
No output schemas documented. LLMs cannot predict response structure, forcing them to guess field names for downstream tool calls. E.g., does get_last_value return {value, timestamp, entity_id} or {data: {value, ts}}?
query_flux tool accepts arbitrary Flux strings with no validation guidance. Description does not warn about injection risk or mention flux_escape() mitigation. LLMs may pass unescaped user input, bypassing server-side protection.
Numeric parameters (hours, limit, sigma, aggregate_minutes) lack min/max constraints. LLMs can pass absurd values (hours=-1, limit=999999) that break queries or cause timeouts. No enum or range validation in schema.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | C | 63 | 2026-07-28+ | v2 |
Arbitrary Flux query against InfluxDB. Returns all tags including entity_id.
Write a single value point to InfluxDB for a sensor. measurement is the schema's unit bucket (e.g. '%', 'kWh', 'W'), not the entity_id -- if unsure which measurement a sensor uses, check list_measurements or get_last_value first.
list_measurements and list_fields return unbounded lists with no pagination. Large buckets could return thousands of items, exhausting context window. No limit parameter or total_count in response.
Error responses not documented. No guidance on retryability, user-fixable vs fatal errors, or recovery steps. E.g., if entity_id is invalid, does the tool return 400/404/500? What should the LLM do next?