MCP Schema Update - October 2025

Marketing Analytics & Geographic Hierarchy Expansion

Date: October 6, 2025 Schema Version: 2.0 Previous Version: 1.0 (County-only, 29 variables) Status: 📝 Documentation Complete - Implementation Pending

🎯 Overview

This document describes the database schema updates for Claude’s MCP (Model Context Protocol) server to support the expanded marketing analytics and 4-level geographic hierarchy.

What Changed

Previous Schema (v1.0):

1 table: county_data
3,144 counties
29 variables (healthcare SDOH focus)
Single geographic level

New Schema (v2.0):

5 tables: state_data, county_data, tract_data, block_group_data_expanded, geo_hierarchy
327,337 total geographies
84 variables (marketing + healthcare)
4-level geographic hierarchy

📊 Database Schema

Table 1: `state_data` (NEW)

Purpose: State-level aggregated demographics Records: 52 (50 states + DC + PR) Primary Key: geoid (2-digit state FIPS)

Schema:

CREATE TABLE state_data (
  -- Geographic identifiers
  state_fips VARCHAR,
  geoid VARCHAR PRIMARY KEY,  -- 2-digit (e.g., "06" for California)
  state_name VARCHAR,

  -- Demographics (16 variables)
  population BIGINT,
  median_age DOUBLE,
  male_population INTEGER,
  female_population INTEGER,
  age_under_5 INTEGER,
  age_5_to_17 INTEGER,
  age_18_to_64 INTEGER,
  age_65_plus INTEGER,
  white_alone INTEGER,
  black_alone INTEGER,
  asian_alone INTEGER,
  hispanic_latino INTEGER,

  -- Economics (16 variables)
  median_household_income INTEGER,
  per_capita_income INTEGER,
  poverty_rate DOUBLE,
  public_assistance INTEGER,
  snap_benefits INTEGER,
  unemployment_rate DOUBLE,
  income_less_10k INTEGER,
  income_10_to_25k INTEGER,
  income_25_to_50k INTEGER,
  income_50_to_75k INTEGER,
  income_75_to_100k INTEGER,
  income_100_to_150k INTEGER,
  income_150_to_200k INTEGER,
  income_200k_plus INTEGER,

  -- Education (5 variables)
  some_high_school_pct DOUBLE,
  high_school_grad_pct DOUBLE,
  some_college_pct DOUBLE,
  bachelors_plus_pct DOUBLE,

  -- Housing (10 variables)
  housing_units INTEGER,
  vacant_units INTEGER,
  renter_occupied_pct DOUBLE,
  rent_burden_50pct_plus DOUBLE,
  crowded_housing_pct DOUBLE,
  single_family_homes INTEGER,
  mobile_homes INTEGER,
  median_year_built INTEGER,
  median_home_value INTEGER,

  -- Technology (6 variables)
  with_computer_pct DOUBLE,
  with_broadband_pct DOUBLE,
  with_smartphone_pct DOUBLE,
  with_tablet_pct DOUBLE,
  no_internet_pct DOUBLE,

  -- Transportation (10 variables)
  commute_less_10_min INTEGER,
  commute_10_to_19_min INTEGER,
  commute_20_to_29_min INTEGER,
  commute_30_to_44_min INTEGER,
  commute_45_plus_min INTEGER,
  work_from_home INTEGER,
  public_transit_pct DOUBLE,
  no_vehicle_transit_pct DOUBLE,
  no_vehicle_walk_pct DOUBLE,

  -- Occupation (5 variables)
  management_occupations_pct DOUBLE,
  business_finance_occupations_pct DOUBLE,
  healthcare_occupations_pct DOUBLE,
  service_occupations_pct DOUBLE,
  retail_sales_pct DOUBLE,

  -- Healthcare (8 variables)
  uninsured_rate DOUBLE,
  uninsured_under_19 INTEGER,
  uninsured_19_to_64 INTEGER,
  uninsured_65_plus INTEGER,
  disability_rate DOUBLE,
  ambulatory_difficulty_pct DOUBLE,
  independent_living_difficulty_pct DOUBLE,

  -- Language (3 variables)
  limited_english_pct DOUBLE,
  spanish_limited_english_pct DOUBLE,
  asian_limited_english_pct DOUBLE,

  -- Family Structure (5 variables)
  children_with_2_parents_pct DOUBLE,
  children_single_parent_pct DOUBLE,
  single_person_households_pct DOUBLE,
  seniors_living_alone_pct DOUBLE,
  grandparents_responsible_pct DOUBLE
);

Sample Query:

SELECT state_name, population, median_household_income, with_broadband_pct
FROM state_data
WHERE population > 10000000
ORDER BY median_household_income DESC;

Table 2: `county_data` (EXISTING - No Changes)

Purpose: County-level demographics Records: 3,144 counties Primary Key: state + county (composite)

Note: This table remains unchanged for backward compatibility.

Table 3: `tract_data` (NEW)

Purpose: Census tract-level demographics Records: 84,400 tracts Primary Key: geoid (11-digit tract FIPS)

Schema: Same 84 variables as state_data, plus:

CREATE TABLE tract_data (
  -- Geographic identifiers
  state_fips VARCHAR,
  county_fips VARCHAR,
  tract_fips VARCHAR,
  geoid VARCHAR PRIMARY KEY,  -- 11-digit (e.g., "06075014500")
  state_name VARCHAR,
  county_name VARCHAR,
  tract_name VARCHAR,

  -- ... same 84 variables as state_data ...
);

Sample Query:

-- Find high-income tracts in San Francisco County
SELECT geoid, tract_name, median_household_income, income_200k_plus
FROM tract_data
WHERE county_fips = '075' AND state_fips = '06'
  AND median_household_income > 150000
ORDER BY median_household_income DESC;

Table 4: `block_group_data_expanded` (NEW)

Purpose: Block group-level demographics with expanded variables Records: 239,741 block groups Primary Key: geoid (12-digit block group FIPS)

Schema: Same 84 variables as state_data, plus:

CREATE TABLE block_group_data_expanded (
  -- Geographic identifiers
  state_fips VARCHAR,
  county_fips VARCHAR,
  tract_fips VARCHAR,
  block_group VARCHAR,
  geoid VARCHAR PRIMARY KEY,  -- 12-digit (e.g., "060750145001")
  state_name VARCHAR,
  county_name VARCHAR,

  -- ... same 84 variables as state_data ...
);

Sample Query:

-- Find tech-savvy block groups with high broadband and smartphone adoption
SELECT geoid, county_name, population,
       with_broadband_pct, with_smartphone_pct
FROM block_group_data_expanded
WHERE state_fips = '06'
  AND with_broadband_pct > 90
  AND with_smartphone_pct > 85
LIMIT 100;

Table 5: `geo_hierarchy` (NEW)

Purpose: Geographic hierarchy metadata for all levels Records: 327,337 (all geographies) Primary Key: geo_type + geoid (composite)

Schema:

CREATE TABLE geo_hierarchy (
  geo_type VARCHAR NOT NULL,       -- 'state', 'county', 'tract', 'block_group'
  geoid VARCHAR NOT NULL,          -- Full GEOID for this geography
  parent_geoid VARCHAR,            -- GEOID of parent geography (NULL for states)
  geo_name VARCHAR,                -- Human-readable name
  population BIGINT,               -- Total population
  land_area_sqmi DOUBLE,           -- Land area in square miles (optional)
  PRIMARY KEY (geo_type, geoid)
);

CREATE INDEX idx_geo_hierarchy_geoid ON geo_hierarchy(geoid);
CREATE INDEX idx_geo_hierarchy_parent ON geo_hierarchy(parent_geoid);
CREATE INDEX idx_geo_hierarchy_type ON geo_hierarchy(geo_type);

Sample Queries:

1. Find parent geography:

SELECT parent_geoid, geo_name
FROM geo_hierarchy
WHERE geoid = '06075';
-- Returns: parent_geoid='06', geo_name='San Francisco County'

2. Find all children of a geography:

SELECT geoid, geo_name, population
FROM geo_hierarchy
WHERE parent_geoid = '06' AND geo_type = 'county'
ORDER BY population DESC;
-- Returns all California counties

3. Get full hierarchy path (recursive):

WITH RECURSIVE hierarchy AS (
  SELECT geo_type, geoid, parent_geoid, geo_name, 1 as level
  FROM geo_hierarchy
  WHERE geoid = '060750145001'

  UNION ALL

  SELECT h.geo_type, h.geoid, h.parent_geoid, h.geo_name, p.level + 1
  FROM geo_hierarchy h
  INNER JOIN hierarchy p ON h.geoid = p.parent_geoid
)
SELECT * FROM hierarchy ORDER BY level DESC;
-- Returns: State → County → Tract → Block Group

🔧 MCP Tool Updates

Updated Tool: `get_information_schema`

Previous Response:

{
  "tables": ["county_data"],
  "county_data": {
    "columns": ["state", "county", "population", "median_income", ...]
  }
}

New Response:

{
  "tables": [
    "state_data",
    "county_data",
    "tract_data",
    "block_group_data_expanded",
    "geo_hierarchy"
  ],
  "state_data": {
    "description": "State-level demographics (52 states)",
    "primary_key": "geoid",
    "record_count": 52,
    "variable_count": 84,
    "columns": {
      "geoid": "VARCHAR - 2-digit state FIPS (e.g., '06')",
      "state_name": "VARCHAR - State name",
      "population": "BIGINT - Total population",
      "median_household_income": "INTEGER - Median household income",
      "with_broadband_pct": "DOUBLE - % with broadband",
      // ... 79 more columns
    }
  },
  "tract_data": {
    "description": "Census tract demographics (84,400 tracts)",
    "primary_key": "geoid",
    "record_count": 84400,
    "variable_count": 84,
    // ... same 84 columns
  },
  "block_group_data_expanded": {
    "description": "Block group demographics (239,741 block groups)",
    "primary_key": "geoid",
    "record_count": 239741,
    "variable_count": 84,
    // ... same 84 columns
  },
  "geo_hierarchy": {
    "description": "Geographic hierarchy metadata (327,337 geographies)",
    "primary_key": ["geo_type", "geoid"],
    "record_count": 327337,
    "columns": {
      "geo_type": "VARCHAR - 'state', 'county', 'tract', 'block_group'",
      "geoid": "VARCHAR - Full GEOID",
      "parent_geoid": "VARCHAR - Parent GEOID (NULL for states)",
      "geo_name": "VARCHAR - Human-readable name",
      "population": "BIGINT - Total population"
    }
  }
}

New Tool: `get_geography_level_recommendation`

Purpose: Suggests optimal geography level for a query based on intent

Input Schema:

{
  name: "get_geography_level_recommendation",
  description: "Recommends the best geographic level (state/county/tract/block_group) for a natural language query",
  inputSchema: {
    type: "object",
    properties: {
      query: {
        type: "string",
        description: "Natural language query from user"
      },
      estimatedPopulation: {
        type: "number",
        description: "Optional: Estimated population size mentioned in query"
      }
    },
    required: ["query"]
  }
}

Example Request:

{
  "query": "Find affluent neighborhoods in Los Angeles",
  "estimatedPopulation": null
}

Example Response:

{
  "recommended_level": "block_group",
  "recommended_table": "block_group_data_expanded",
  "reasoning": "Query asks for 'neighborhoods' which suggests granular block group level. Los Angeles is a large county, so block group provides necessary detail.",
  "estimated_rows": 5000,
  "alternative_levels": [
    {
      "level": "tract",
      "table": "tract_data",
      "reasoning": "Could use tracts for broader neighborhood analysis",
      "estimated_rows": 2000
    }
  ]
}

New Tool: `get_variable_categories`

Purpose: Returns all 10 variable categories with variable counts and descriptions

Input Schema:

{
  name: "get_variable_categories",
  description: "Lists all variable categories with counts and example variables",
  inputSchema: {
    type: "object",
    properties: {
      category: {
        type: "string",
        description: "Optional: Filter to specific category (demographics, economics, technology, etc.)",
        enum: ["demographics", "economics", "education", "housing", "technology",
               "transportation", "occupation", "healthcare", "language", "family"]
      }
    }
  }
}

Example Response (all categories):

{
  "categories": [
    {
      "name": "demographics",
      "variable_count": 16,
      "description": "Population, age, race, ethnicity",
      "example_variables": [
        "population", "median_age", "age_65_plus", "hispanic_latino"
      ],
      "use_cases": ["Market sizing", "Audience targeting", "Demographic segmentation"]
    },
    {
      "name": "economics",
      "variable_count": 16,
      "description": "Income distribution, poverty, assistance programs",
      "example_variables": [
        "median_household_income", "income_200k_plus", "poverty_rate", "snap_benefits"
      ],
      "use_cases": ["Pricing strategy", "Affordability analysis", "Premium targeting"]
    },
    {
      "name": "technology",
      "variable_count": 6,
      "description": "Digital access, device ownership, internet connectivity",
      "example_variables": [
        "with_broadband_pct", "with_smartphone_pct", "no_internet_pct"
      ],
      "use_cases": ["E-commerce readiness", "Digital marketing", "Tech product targeting"]
    },
    // ... 7 more categories
  ],
  "total_variables": 84
}

New Tool: `get_hierarchy_path`

Purpose: Gets full geographic hierarchy for any location

Input Schema:

{
  name: "get_hierarchy_path",
  description: "Returns the full geographic hierarchy (state → county → tract → block group) for a given GEOID",
  inputSchema: {
    type: "object",
    properties: {
      geoid: {
        type: "string",
        description: "GEOID at any level (2-digit state, 5-digit county, 11-digit tract, 12-digit block group)"
      },
      include_siblings: {
        type: "boolean",
        description: "Include sibling geographies at each level (default: false)"
      }
    },
    required: ["geoid"]
  }
}

Example Request:

{
  "geoid": "060750145001",
  "include_siblings": false
}

Example Response:

{
  "full_path": [
    {
      "level": "state",
      "geoid": "06",
      "name": "California",
      "population": 39242785
    },
    {
      "level": "county",
      "geoid": "06075",
      "name": "San Francisco County",
      "population": 851036
    },
    {
      "level": "tract",
      "geoid": "06075014500",
      "name": "Tract 145; San Francisco County; California",
      "population": 4521
    },
    {
      "level": "block_group",
      "geoid": "060750145001",
      "name": "Block Group 1",
      "population": 1205
    }
  ],
  "total_levels": 4
}

🔒 Security Policy Updates

Updated: `sqlSecurityPolicies.ts`

Add to Table Allowlist:

const ALLOWED_TABLES = [
  'county_data',              // existing
  'state_data',               // NEW
  'tract_data',               // NEW
  'block_group_data_expanded', // NEW
  'geo_hierarchy'             // NEW
];

Add to Column Validation:

const ALLOWED_COLUMNS = {
  state_data: ['*'],  // Allow all 84 + geographic columns
  tract_data: ['*'],
  block_group_data_expanded: ['*'],
  geo_hierarchy: [
    'geo_type', 'geoid', 'parent_geoid', 'geo_name', 'population', 'land_area_sqmi'
  ],
  county_data: ['*']  // existing
};

Row Limit: Keep at 1,000 max per query (prevents performance issues with 240K block groups)

📝 Claude System Prompt Updates

Update: `anthropicService.ts` Schema Description

Add to System Prompt:

## Database Schema

You have access to 5 tables with comprehensive U.S. Census demographic data:

### Geographic Hierarchy (4 Levels)

1. **state_data** (52 states)
   - GEOID format: 2-digit (e.g., "06" = California)
   - Use for: State comparisons, national trends

2. **county_data** (3,144 counties)
   - GEOID format: 5-digit (e.g., "06075" = San Francisco County)
   - Use for: Regional analysis, market studies

3. **tract_data** (84,400 census tracts)
   - GEOID format: 11-digit (e.g., "06075014500")
   - Avg population: 4,500
   - Use for: Neighborhood analysis, local targeting

4. **block_group_data_expanded** (239,741 block groups)
   - GEOID format: 12-digit (e.g., "060750145001")
   - Avg population: 1,400
   - Use for: Micro-targeting, hyperlocal analysis

5. **geo_hierarchy** (327,337 geographies)
   - Links all levels with parent-child relationships
   - Use for: Hierarchical queries, geographic navigation

### 84 Variables Across 10 Categories

**Demographics (16 variables)**:
- population, median_age, male_population, female_population
- age_under_5, age_5_to_17, age_18_to_64, age_65_plus
- white_alone, black_alone, asian_alone, hispanic_latino

**Economics (16 variables)**:
- median_household_income, per_capita_income, poverty_rate
- income_less_10k through income_200k_plus (8 income brackets)
- unemployment_rate, public_assistance, snap_benefits

**Education (5 variables)**:
- some_high_school_pct, high_school_grad_pct
- some_college_pct, bachelors_plus_pct

**Housing (10 variables)**:
- housing_units, vacant_units, renter_occupied_pct
- rent_burden_50pct_plus, crowded_housing_pct
- single_family_homes, mobile_homes, median_year_built, median_home_value

**Technology & Digital Access (6 variables)**:
- with_computer_pct, with_broadband_pct, with_smartphone_pct
- with_tablet_pct, no_internet_pct

**Transportation & Commuting (10 variables)**:
- commute_less_10_min, commute_10_to_19_min, etc. (5 time brackets)
- work_from_home, public_transit_pct
- no_vehicle_transit_pct, no_vehicle_walk_pct

**Occupation & Industry (5 variables)**:
- management_occupations_pct, business_finance_occupations_pct
- healthcare_occupations_pct, service_occupations_pct, retail_sales_pct

**Healthcare & Insurance (8 variables)**:
- uninsured_rate, uninsured_under_19, uninsured_19_to_64, uninsured_65_plus
- disability_rate, ambulatory_difficulty_pct, independent_living_difficulty_pct

**Language & Cultural Access (3 variables)**:
- limited_english_pct, spanish_limited_english_pct, asian_limited_english_pct

**Family Structure & Social (5 variables)**:
- children_with_2_parents_pct, children_single_parent_pct
- single_person_households_pct, seniors_living_alone_pct
- grandparents_responsible_pct

### Query Guidelines

**Geographic Level Selection**:
- State comparisons → `state_data`
- Market/county analysis → `county_data`
- Neighborhood targeting → `tract_data`
- Micro-targeting → `block_group_data_expanded`

**Performance Optimization**:
- ALWAYS filter by geography first (state_fips, county_fips)
- Use appropriate level (don't aggregate from block groups unnecessarily)
- Apply LIMIT for large result sets

**Hierarchical Queries**:
- Use `geo_hierarchy` for parent-child lookups
- Substring GEOIDs for filtering:
  - State: SUBSTRING(geoid, 1, 2)
  - County: SUBSTRING(geoid, 1, 5)
  - Tract: SUBSTRING(geoid, 1, 11)

### Example Queries

1. **State Comparison**:
   ```sql
   SELECT state_name, median_household_income, with_broadband_pct
   FROM state_data
   WHERE population > 5000000
   ORDER BY median_household_income DESC;

Affluent Neighborhoods:

SELECT geoid, county_name, income_200k_plus, median_household_income
FROM block_group_data_expanded
WHERE state_fips = '06' AND income_200k_plus > 100
ORDER BY income_200k_plus DESC LIMIT 50;

Tech-Ready Markets:

SELECT county_name, AVG(with_broadband_pct) as avg_broadband
FROM tract_data
WHERE state_fips = '06'
GROUP BY county_name
HAVING AVG(with_broadband_pct) > 85;

Hierarchy Navigation:

SELECT geoid, geo_name, population
FROM geo_hierarchy
WHERE parent_geoid = '06' AND geo_type = 'county'
ORDER BY population DESC;

```

🧪 Testing Scenarios

Natural Language Query Examples

Geography Level Routing:

State Level:
- “Compare California and Texas by median income” → state_data
- “Which states have the highest broadband adoption?” → state_data
County Level:
- “Show me demographics for San Francisco County” → county_data
- “Find counties in California with poverty rate > 15%” → county_data (or join state)
Tract Level:
- “What are the demographics of neighborhoods in downtown LA?” → tract_data
- “Find tracts with high tech worker concentration” → tract_data
Block Group Level:
- “Target affluent neighborhoods in Silicon Valley” → block_group_data_expanded
- “Find micro-areas with high senior populations” → block_group_data_expanded

Variable Category Coverage:

Marketing Analytics:
- “Where are high-income consumers located?” → Uses income_200k_plus
- “Find areas with long commutes” → Uses commute_45_plus_min
- “Which neighborhoods have high broadband access?” → Uses with_broadband_pct
Healthcare Marketing:
- “Find uninsured working-age adults” → Uses uninsured_19_to_64
- “Where are Spanish-speaking communities with limited English?” → Uses spanish_limited_english_pct
- “Target isolated seniors for care services” → Uses seniors_living_alone_pct

Cross-Geography Queries:

Hierarchical Rollup:
- “Show me California counties ranked by broadband access” → Aggregate from tract_data or use county_data
Drill-Down:
- “Find block groups within San Francisco County with >$150k income” → Filter block_group_data_expanded by county

🚀 Implementation Checklist

Phase 1: Schema Updates ✅ COMPLETE

Load state_data (52 states)
Load tract_data (84,400 tracts)
Load block_group_data_expanded (239,741 block groups)
Create geo_hierarchy (327,337 geographies)

Phase 2: MCP Server Updates (NEXT)

Update get_information_schema tool response
Implement get_geography_level_recommendation tool
Implement get_variable_categories tool
Implement get_hierarchy_path tool
Update anthropicService.ts system prompt
Update sqlSecurityPolicies.ts table allowlist

Phase 3: Testing & Validation

Test all MCP tools with new schema
Run natural language query eval suite
Verify geography level routing accuracy
Validate SQL generation for all tables
Performance test with block group queries

Phase 4: Documentation & Training

Update API documentation
Create user guide for new variables
Document performance best practices
Create video tutorials (optional)

📊 Performance Considerations

Query Performance Guidelines

Table	Records	Unfiltered Query	With State Filter	With County Filter
state_data	52	<1ms	N/A	N/A
county_data	3,144	~10ms	~5ms	N/A
tract_data	84,400	~100ms	~20ms	~5ms
block_group_data_expanded	239,741	~500ms	~50ms	~10ms

Optimization Strategies:

Always Filter by Geography:

-- GOOD
SELECT * FROM block_group_data_expanded
WHERE state_fips = '06' AND median_household_income > 100000;

-- BAD (scans 240K rows)
SELECT * FROM block_group_data_expanded
WHERE median_household_income > 100000;

Use Appropriate Level:

-- GOOD: State comparison
SELECT state_name, AVG(poverty_rate) FROM state_data;

-- BAD: Unnecessary aggregation
SELECT state_name, AVG(poverty_rate)
FROM block_group_data_expanded GROUP BY state_name;

Create Indexes (recommended):

CREATE INDEX idx_blockgroup_state ON block_group_data_expanded(state_fips);
CREATE INDEX idx_blockgroup_income ON block_group_data_expanded(median_household_income);
CREATE INDEX idx_tract_poverty ON tract_data(poverty_rate);

Main Expansion Guide: /docs/MARKETING_ANALYTICS_EXPANSION.md
Quick Start: /docs/QUICK_START_EXPANDED.md
Epic 3 Documentation: docs/internal/epics/epic-3-marketing-analytics-expansion.md
Session Summary: docs/internal/sessions/SESSION_OCT_6_2025_EXPANSION.md
MCP API Docs: /docs/api/MCP_API_DOCUMENTATION.md

Document Status: ✅ Complete - Ready for Implementation Next Step: Update anthropicService.ts and mcpServer.ts with new schema Estimated Implementation Time: 2-3 hours

Last Updated: October 6, 2025 Schema Version: 2.0 Author: Kevin Kellogg

MCP Schema Update - October 2025

Marketing Analytics & Geographic Hierarchy Expansion

🎯 Overview

What Changed

📊 Database Schema

Table 1: state_data (NEW)

Table 2: county_data (EXISTING - No Changes)

Table 3: tract_data (NEW)

Table 4: block_group_data_expanded (NEW)

Table 5: geo_hierarchy (NEW)

🔧 MCP Tool Updates

Updated Tool: get_information_schema

New Tool: get_geography_level_recommendation

New Tool: get_variable_categories

New Tool: get_hierarchy_path

🔒 Security Policy Updates

Updated: sqlSecurityPolicies.ts

📝 Claude System Prompt Updates

Update: anthropicService.ts Schema Description

🧪 Testing Scenarios

Natural Language Query Examples

🚀 Implementation Checklist

Phase 1: Schema Updates ✅ COMPLETE

Phase 2: MCP Server Updates (NEXT)

Phase 3: Testing & Validation

Phase 4: Documentation & Training

📊 Performance Considerations

Query Performance Guidelines

🔗 Related Documentation

Table 1: `state_data` (NEW)

Table 2: `county_data` (EXISTING - No Changes)

Table 3: `tract_data` (NEW)

Table 4: `block_group_data_expanded` (NEW)

Table 5: `geo_hierarchy` (NEW)

Updated Tool: `get_information_schema`

New Tool: `get_geography_level_recommendation`

New Tool: `get_variable_categories`

New Tool: `get_hierarchy_path`

Updated: `sqlSecurityPolicies.ts`

Update: `anthropicService.ts` Schema Description