ThinAir Data

detect_anomalies

Read-onlyIdempotent

Scan a table for unusual patterns: volume drops/spikes, data gaps, value concentration, high null rates, stale data. Severity-ranked alerts. Tables > 100k rows use a sampled path (~5%) — when a finding has sampled:true, surface it to the user with a hedge like 'based on a ~5% sample' rather than presenting the number as exact. Dialect-aware: TABLESAMPLE SYSTEM on postgres, TABLESAMPLE PERCENT on mssql, WHERE RAND() on mysql.

Input Schema

TableJSON Schema

Name	Required	Description
`table`	Yes	Table to scan for anomalies
`connection`	No	Target connection name from this tenant's inventory. Call `list_connections` to see every name + dialect, then match semantically to the user's intent (e.g. 'analytics' → a connection named `-analytics-`; 'prod' → a connection with `prod-` prefix). If the user didn't specify, use the tenant's default (first added). Do not invent names — resolve from `list_connections` output.
`date_column`	No	Date column for trend analysis (auto-detected if omitted)

Output Schema

TableJSON Schema

Name	Required	Description	Default
`data`	No
`meta`	No
`display`	No
`summary`	No
`insights`	No

TDQS

A4.5/5.0

Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations already declare readOnly, openWorld, and idempotent hints, but the description adds critical behavioral details: tables >100k rows are sampled at ~5%, results may carry 'sampled:true', and the exact sampling mechanism varies by SQL dialect. This goes well beyond what annotations convey and helps the agent set user expectations correctly.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is concise and well-structured: it opens with purpose and anomaly types, mentions severity-ranked alerts, then explains sampling and dialect behavior. Every sentence carries useful information with no redundancy.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Given an output schema exists and parameter descriptions are comprehensive, the description completes the picture by covering sampling caveats, dialect-specific behavior, and how to present sampled results. It is sufficiently complete for correct invocation and result interpretation.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The input schema already provides full descriptions for all three parameters, including rich guidance for the 'connection' parameter. The description adds no additional parameter-level meaning beyond what the schema covers, so the baseline score of 3 applies.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states the tool scans a table for unusual patterns and enumerates specific anomaly types (volume drops/spikes, data gaps, value concentration, high null rates, stale data). This specific verb+resource combination differentiates it from generic analysis tools like data_profile or analyze_table.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description gives strong usage context by describing the sampling behavior for large tables and how to report sampled results, plus dialect-specific SQL methods. However, it does not explicitly name alternative tools or state when not to use this tool, so it falls short of a perfect score.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

All ThinAir Data tools Inspect

Try in Browser

RayrunBuild, deploy, and operate MCP servers without becoming an infrastructure team.ray.runAd

Glama MCP Gateway

Add one secure layer between your agents and this server.

TDQS

A3.9/5.0

Disambiguation4/5

Each tool targets a distinct resource/action, though some overlap exists between analyze_table and data_profile (both profile tables) and between pii_scan and configure_allowlist (both deal with PII). Clear descriptions and separation of quick vs. full analysis mitigate confusion.

Naming Consistency3/5

Most tools follow verb_noun (add_connection, describe_schema, generate_migration), but a notable minority use noun phrases (data_profile, pii_scan, query_firewall, saved_queries, quota, impact_analysis). This mixed convention creates inconsistency, though the naming is still readable.

Tool Count3/5

26 tools is slightly over the 16-25 heavy threshold, but each tool addresses a distinct need across connection management, querying, analysis, security, and performance. While the count feels high, the breadth justifies it; however, it's approaching the 'too many' range.

Completeness5/5

The tool set covers the full lifecycle: connections (add/remove/list/test), querying (query_sql, saved_queries, cross_db_query), schema exploration/migration (describe_schema, generate_migration, impact_analysis), data quality/compliance (analyze_table, data_profile, pii_scan), performance (explain_query, optimize_query, show_locks), and monitoring (watch_table, detect_anomalies). Any gaps are minor, such as no update_connection, but that's not a core need.

Resources

Need Help?