Meerkats Tables MCP Server

This MCP (Model Context Protocol) server provides programmatic access to Meerkats tables using API key authentication.

Features

• List Tables: Get all your tables with pagination and search
• CRUD Operations: Create, read, update, and delete tables
• Row Management: Add, update, and delete rows in your tables
• Statistics: Get table usage stats
• API Key Authentication: Secure access using your Meerkats API keys

Setup

1. Get API Key:

   • Log in to your Meerkats dashboard
   • Go to https://app.meerkats.ai/dashboard/api-keys
   • Create a new API key with table permissions

2. Claude Desktop Configuration: Add this to your Claude Desktop claude_desktop_config.json:

   {
     "mcpServers": {
       "meerkats": {
         "command": "npx",
         "args": ["-y", "meerkats"],
         "env": {
           "API_KEY": "your-api-key-here"
         }
       }
     }
   }

Usage

Available Tools

Core Table Management

1. list_tables - List all your tables with pagination support
2. get_table - Get details of a specific table
3. create_table - Create a new table
4. update_table - Update an existing table
5. delete_table - Delete a table

Table Row Operations

6. get_table_rows - Get rows from a table with pagination support
7. add_table_row - Add a new row to a table
8. add_table_rows_bulk - Add multiple rows to a table in bulk
9. update_table_row - Update an existing row
10. delete_table_row - Delete a row from a table
11. filter_table_rows - Filter and search rows using column filters and row ranges

Table Column Management

12. add_table_column - Add a new column to an existing table
13. update_table_column - Update an existing column in a table
14. schedule_table_column - Schedule an AI column to run automatically
15. get_table_column_schedule - Get the current scheduling information for an AI column

Sheet Management (Sub-tables)

16. list_table_sheets - List all sheets within a table with pagination support
17. get_table_sheet - Get details of a specific sheet by ID
18. create_table_sheet - Create a new sheet within a table
19. update_table_sheet - Update an existing sheet
20. delete_table_sheet - Delete a sheet from a table

Sheet Row Operations

21. get_table_sheet_rows - Get rows from a specific sheet with pagination support
22. add_table_sheet_row - Add a new row to a sheet
23. update_table_sheet_row - Update an existing row in a sheet
24. delete_table_sheet_row - Delete a row from a sheet
25. filter_table_sheet_rows - Filter and search rows in a sheet using column filters and row range

Utilities

26. get_table_stats - Get table statistics
27. get_mcp_servers - Get available MCP servers and their tools

External Tools

28. tavily_search - Search the web using Tavily search engine
29. google_maps_search - Search for locations using Google Maps API (location search only)
30. web_scrape - Scrape content from a single web URL

AI Cell Processing

31. run_table_ai_cell - Run AI processing for a single table cell
32. run_table_ai_cells_bulk - Run AI processing for multiple table cells in bulk

Data Visualization

33. generate_vibe_visualization - Generate dynamic interactive visualizations using vibe coding with intelligent data analysis and custom D3.js adaptations
34. create_universal_visualization - Create any type of visualization directly from raw data using advanced vibe coding - no table creation required

Sheet System (Sub-tables)

Tables can contain sheets, which are sub-tables with the same structure as the parent table but with additional organization capabilities.

Tool Reference System

When creating AI columns, you can specify MCP tools that the column should use for processing. Tools are referenced using the format server_name.tool_name:

How It Works:

1. Tool Format: Use "server_name.tool_name" format (e.g., "github.search_repositories")
2. Backend Resolution: Backend automatically finds the MCP server and tool by name
3. Tool Attachment: Resolved tools are attached to the column for AI processing

Example Tool References:

• "web.fetch_page" - Web scraping tool
• "github.search_repositories" - GitHub search tool
• "finance.get_company_data" - Financial data tool
• "openapi.call_endpoint" - API calling tool

Getting Available Tools:

Use the get_mcp_servers tool to see all available MCP servers and their tools in your system.

Row Filtering System

The filter_table_rows tool provides powerful filtering capabilities for table data using the same filtering logic as the Meerkats UI:

Filter Structure

Each filter consists of:

• column: Column metadata (id, label, datatype)
• operator: Filter operator (==, !=, >, <, contains, empty, etc.)
• value: Value to filter by (not required for empty/notempty operators)
• condition: How to combine with other filters (And/Or)

Supported Operators by Data Type

Text Columns:

• == - Equals
• != - Not equals
• contains - Contains text
• does not contain - Does not contain text
• empty - Field is empty
• notempty - Field is not empty

Number Columns:

• == - Equals
• != - Not equals
• > - Greater than
• < - Less than
• >= - Greater than or equal
• <= - Less than or equal
• empty - Field is empty
• notempty - Field is not empty

Date Columns:

• == - Equals date
• != - Not equals date
• > - After date
• < - Before date
• >= - On or after date
• <= - On or before date
• empty - Field is empty
• notempty - Field is not empty

Special AI Status Operators:

• error - Shows rows with errors
• notfound - Shows "not found" results
• inqueue - Shows queued tasks
• processing - Shows processing tasks

Row Range Filtering

• from: Start from specific row number (1-based)
• to: End at specific row number (1-based)
• Can be combined with column filters

Pagination

• limit: Maximum rows to return (default: 50)
• page: Page number for pagination (default: 0)

Example Filter Scenarios

AI Column Scheduling

AI columns can be scheduled to run automatically using cron expressions or single-run schedules:

Schedule Types:

1. Recurring Schedule: Uses cron expressions for repeated execution

   • "0 9 * * *" - Daily at 9 AM
   • "0 */6 * * *" - Every 6 hours
   • "0 9 * * 1" - Weekly on Monday at 9 AM
   • "0 9 1 * *" - Monthly on 1st at 9 AM

2. Single Run: Execute once at a specific time

   • Uses ISO datetime format: "2024-01-15T10:30:00Z"

Scheduling Features:

• AI Column Only: Only AI columns can be scheduled (Input columns cannot)
• Job Management: Each scheduled column gets a unique job ID
• Run Tracking: Tracks last run and next run times
• Enable/Disable: Easily enable or disable scheduling without losing configuration

Common Cron Patterns:

• "0 * * * *" - Every hour
• "*/15 * * * *" - Every 15 minutes
• "0 8,12,18 * * *" - Three times daily (8 AM, 12 PM, 6 PM)
• "0 9 * * 1-5" - Weekdays at 9 AM

Dynamic Data Visualization with Vibe Coding

The MCP server features advanced "vibe coding" capabilities that dynamically generate interactive visualizations at runtime. Unlike static chart libraries, this system analyzes your data characteristics and creates custom D3.js-powered visualizations with intelligent adaptations, animations, and real-time capabilities.

Chart Types Supported

The visualization tools support 9 different chart types:

• Bar Charts (bar) - Great for comparing categories
• Line Charts (line) - Perfect for trends over time
• Pie Charts (pie) - Ideal for showing proportions
• Doughnut Charts (doughnut) - Modern alternative to pie charts
• Scatter Plots (scatter) - For correlations between variables
• Area Charts (area) - Filled line charts for cumulative data
• Histograms (histogram) - For data distribution analysis
• Heatmaps (heatmap) - For matrix/correlation data
• Treemaps (treemap) - For hierarchical data visualization

Vibe Coding Features

🎨 Intelligent Code Generation:

• Analyzes data types (numeric, categorical, temporal, text)
• Suggests optimal chart types and column mappings
• Generates adaptive D3.js code based on data characteristics
• Creates custom animations and interactions

⚡ Runtime Adaptability:

• Dynamic scale adjustments based on data ranges
• Smart margin calculations for labels
• Responsive layouts that adapt to content
• Theme-aware styling (light/dark mode)

✨ Interactive Elements:

• Hover effects and tooltips
• Animated transitions (bars growing, lines drawing)
• Click interactions and data exploration
• Real-time data refresh capabilities

Configuration

Environment Variables

• API_KEY - Your Meerkats API key (set in Claude Desktop config)

API Key Permissions

Your API key needs the following permissions for full functionality:

• table:list - List and view tables
• table:create - Create new tables
• table:update - Update existing tables
• table:delete - Delete tables
• read - General read access
• write - General write access
• delete - General delete access

Running the Server

npm start

The server will start and listen for MCP protocol messages on stdin/stdout.

Integration

This MCP server can be integrated with:

• Claude Desktop
• Other MCP-compatible applications
• Custom applications using the MCP protocol

Support

For issues with the MCP server, check:

1. Backend server is running and accessible
2. API key is valid and has correct permissions
3. Environment configuration is correct
4. Network connectivity between MCP server and backend