Looker Analyst

Quick Start

Using Looker MCP Tools

The Looker MCP provides programmatic access to your Looker instance. Start every workflow by discovering available models and explores:

1. Call mcp__looker-toolbox__get_models to list available models
2. Call mcp__looker-toolbox__get_explores with a model to see available explores
3. Call mcp__looker-toolbox__get_dimensions and get_measures to discover fields
4. Build queries using the query tool

Basic Query Pattern

// Standard query structure
{
  "model": "model_name",           // From get_models
  "explore": "explore_name",       // From get_explores
  "fields": [                      // From get_dimensions/get_measures
    "dimension.field_name",
    "measure.field_name"
  ],
  "filters": {                     // Optional
    "dimension.field": "value"
  },
  "sorts": ["field desc 0"],       // Optional
  "limit": 500                     // Optional, default 500
}

Core Workflows

1. Exploratory Data Analysis

Goal: Understand data structure and discover insights

Steps:

1. Discover structure:

   • get_models → List available data models
   • get_explores → Find relevant explores
   • get_dimensions + get_measures → See available fields

2. Build initial query:

   • Start with key dimensions (grouping fields)
   • Add relevant measures (aggregations)
   • Use query tool to execute

3. Refine and filter:

   • Apply filters to narrow scope
   • Add sorts to prioritize insights
   • Adjust limit for appropriate data size

4. Analyze results:

   • Look for patterns, outliers, trends
   • Identify follow-up questions
   • Iterate with new queries

2. Creating Visualizations

Goal: Build charts and dashboards

Steps:

1. Start with a query (see workflow above)

2. Choose visualization type:

   • Line charts: Time series trends
   • Bar/Column: Comparisons across categories
   • Pie/Donut: Proportional breakdowns
   • Scatter: Correlations between metrics
   • Table: Detailed data review

3. Generate URL or create Look:

   • query_url tool: Get shareable Looker URL with vis_config
   • make_look tool: Save as permanent Look in personal folder

4. Add to dashboard:

   • Use add_dashboard_element to add tiles
   • Configure vis_config for visual styling
   • See reference/visualizations.md for vis_config details

3. Building Dashboards

Goal: Create comprehensive dashboard with multiple tiles

Steps:

1. Plan dashboard structure:

   • Identify key metrics and KPIs
   • Group related visualizations
   • Consider filter dependencies

2. Create dashboard tiles:

   • Build query for each tile
   • Choose appropriate visualization
   • Use add_dashboard_element repeatedly

3. Configure interactivity:

   • Set up dashboard filters
   • Enable cross-filtering if needed
   • Test drill-down paths

4. SQL Generation

Goal: See underlying SQL for debugging or learning

Steps:

1. Build query using normal parameters
2. Use query_sql tool instead of query
3. Review generated SQL
4. Optimize or troubleshoot as needed

5. Health Monitoring

Goal: Monitor Looker instance health and performance

Available checks:

• health_pulse with actions:

  • check_db_connections: Verify database connectivity
  • check_dashboard_performance: Find slow dashboards
  • check_dashboard_errors: Identify broken dashboards
  • check_explore_performance: Find slow explores
  • check_schedule_failures: Check scheduled delivery failures
  • check_legacy_features: Identify deprecated features

• health_analyze: Calculate usage metrics

  • Actions: projects, models, explores
  • Parameters: timeframe (days), min_queries

• health_vacuum: Suggest unused resources for cleanup

  • Actions: models, explores
  • Find candidates for deletion based on usage

Looker MCP Integration

Available MCP Tools

The Looker MCP server (looker-toolbox) provides these tools:

Discovery Tools

• get_models - List all LookML models
• get_explores - List explores in a model
• get_dimensions - List dimensions in an explore
• get_measures - List measures in an explore
• get_filters - List available filters
• get_parameters - List parameters

Query Tools

• query - Execute query and get JSON results
• query_sql - Generate SQL without executing
• query_url - Generate Looker URL with optional vis_config

Content Creation

• make_look - Create saved Look in personal folder
• add_dashboard_element - Add tile to dashboard
• run_look - Execute existing Look by ID
• get_looks - Search for Looks

Development Tools

• get_projects - List LookML projects
• dev_mode - Switch to/from development mode
• create_project_file - Create LookML file
• update_project_file - Modify LookML file
• delete_project_file - Remove LookML file

Health & Monitoring

• health_pulse - Run health checks
• health_analyze - Analyze usage patterns
• health_vacuum - Find unused resources

Tool Call Pattern

All MCP tools are called with this naming convention:

mcp__looker-toolbox__<tool_name>

Example:

mcp__looker-toolbox__query
mcp__looker-toolbox__get_models
mcp__looker-toolbox__make_look

Filter Expressions

Filters use special syntax - see reference/filter-expressions.md for complete syntax.

Quick examples:

{
  "filters": {
    "orders.status": "completed,pending",     // Multiple values (OR)
    "orders.created_date": "7 days",          // Relative date
    "users.age": ">=21 AND <=65",             // Number range
    "products.name": "%shirt%",               // Contains (wildcard)
    "orders.total": "NOT NULL"                // Not null
  }
}

Important: Use not null (not -NULL) for null checks.

When to Use This Skill

Invoke this skill when:

• Building Looker queries programmatically
• Creating visualizations or dashboards
• Analyzing data through Looker explores
• Generating SQL from Looker models
• Creating or modifying Looks
• Working with LookML development
• Monitoring Looker instance health
• Troubleshooting query performance
• Understanding Looker data structure
• Teaching users how to use Looker

Reference Materials

For detailed information, see these reference files (loaded on-demand):

• reference/explores-guide.md - Complete guide to using explores interactively
• reference/filtering.md - All filter types, operators, and expressions
• reference/expressions-and-calcs.md - Expressions, table calculations, and custom fields
• reference/functions-reference.md - Complete function library with examples
• reference/formatting-guide.md - Number and time formatting patterns
• reference/visualizations-guide.md - Chart types and vis_config options
• reference/merged-results-guide.md - Merging results from multiple explores

These files are loaded on-demand when you need specific details.

Key Principles

1. Start with discovery: Always use get_models, get_explores, and field discovery tools first
2. Build iteratively: Start simple, add complexity as needed
3. Validate fields: Use get_dimensions/get_measures to confirm field names
4. Check filter syntax: Reference filter expressions guide for correct syntax
5. Test queries: Use query tool before creating permanent Looks
6. Monitor performance: Use health tools to identify issues early
7. Document decisions: Explain query logic and visualization choices

Common Patterns

Pattern: Multi-Step Analysis

1. get_models → find relevant model
2. get_explores → find relevant explore
3. get_dimensions + get_measures → discover fields
4. query with simple fields → validate data
5. query with filters/sorts → refine results
6. query_url with vis_config → create visualization
7. make_look → save for reuse

Pattern: Dashboard Creation

1. Plan: Identify 4-6 key metrics
2. For each metric:
   - Build query
   - Choose visualization
   - call add_dashboard_element
3. Test interactivity
4. Share dashboard URL

Pattern: Performance Troubleshooting

1. health_pulse check_explore_performance → find slow explores
2. query_sql → examine generated SQL
3. Optimize: Reduce fields, add filters, check joins
4. query → validate improvement

Getting Help

For specific topics, tell me what you need:

• "How do I use explores?" → loads explores-guide.md
• "Show me filter syntax" → loads filtering.md
• "How do I create table calculations?" → loads expressions-and-calcs.md
• "What functions are available?" → loads functions-reference.md
• "How do I format numbers/dates?" → loads formatting-guide.md
• "What chart types can I use?" → loads visualizations-guide.md
• "How do I merge results?" → loads merged-results-guide.md

I'll load the relevant reference material and guide you through the task.

##Examples

See the examples/ directory for common patterns:

• Query building workflows
• Dashboard creation examples
• Complex filter patterns
• Advanced calculation techniques