Skip to main content

Explore Data with AI

Connect our PortalJS MCP Server to ChatGPT or Claude Desktop to search, explore, and preview datasets using natural language.

Setup Instructions

Connecting to ChatGPT

  1. Open ChatGPT
  2. Go to Settings → Model Context Protocol (MCP)
  3. Click Add MCP Server
  4. Enter the following details:
    • Name: PortalJS Data Portal
    • URL: 
      https://mcp.portaljs.com/sse
    • Type: SSE (Server-Sent Events)
  5. Click Save
  6. The server should now appear in your MCP connections list

Connecting to Claude Desktop

  1. Open Claude Desktop
  2. Go to Settings → Connectors
  3. Click Add Custom Connector
  4. Enter the connector URL: 
    https://mcp.portaljs.com/sse
  5. Click Add
  6. The PortalJS server should now be available in your MCP connections

How to Use

Once connected, you can interact with the data portal using natural language. The AI will automatically use the appropriate tools based on your requests.

Search for Datasets

Find datasets by describing what you're looking for:

Example prompts:

  • "Search for datasets about CO2 emissions"
  • "Find data on housing prices"
  • "Look for climate change datasets"
  • "Show me datasets related to population statistics"

The search will return a list of matching datasets with their titles, descriptions, and IDs.

Fetch Dataset Details

Get detailed information about a specific dataset:

Example prompts:

  • "Fetch details for dataset idb--house_price"
  • "Show me information about the CO2 emissions dataset"
  • "Get the resources for the housing dataset"
  • "What resources are available in dataset xyz-123?"

The fetch tool returns complete dataset metadata including:

  • Title and description
  • Organization
  • Resources (files) with formats and download URLs
  • Tags and creation dates
  • License information

Preview Dataset

View the actual data from any resource in table format:

Example prompts:

  • "Preview this dataset"
  • "Show me the data from resource abc-123"
  • "I want to see a preview of the house pricing data"
  • "Display the first 20 rows of this resource"

The preview displays:

  • Data formatted as a markdown table
  • Download link to the full resource
  • Number of records available
  • Source information (DataStore or direct CSV/JSON)

[NOTE] You can specify the number of rows (default: 10, max: 100):

  • "Preview 50 rows of this dataset"
  • "Show me the first 15 records"

Complete Workflow Example

Here's a typical conversation flow:

You: "Search for datasets about house pricing"

AI: [Returns list of housing datasets]

You: "Fetch details for the first one"

AI: [Shows complete dataset information with resources]

You: "Preview the CSV resource"

AI: [Displays data table with download link]

Available Tools

The MCP server provides three main tools:

  1. search - Search datasets by keywords
  2. fetch - Get detailed dataset information
  3. preview_data_table - View data in table format with download links

All tools work seamlessly through natural language - no need to memorize commands or syntax!