Skip to main content

MCP Integration

Model Context Protocol (MCP) Integration

Connect your AI assistant to Coalesce Catalog and unlock intelligent data discovery through natural conversation.

❓What is MCP?​

The Model Context Protocol (MCP) is an open standard that enables AI assistants to connect to external data sources and tools. Think of it as a universal adapter—your AI assistant can now understand and work with your organization's data catalog in real-time.

Instead of searching through documentation or clicking through UIs, simply ask your AI: "Find all customer tables owned by the {your-team-name} team" and get instant, accurate results.

✨Why Use Coalesce MCP?​

The Coalesce MCP complements the AI search already available in Coalesce Catalog by letting you use your own AI assistant while leveraging the same powerful search capabilities. This means you can:

  • Use Your Preferred AI - Work with Claude, ChatGPT, Cursor, or any MCP-compatible assistant
  • Instant Answers - Get real-time information about tables, owners, and documentation from your catalog
  • Contextual Understanding - Your AI knows your data warehouse structure and business terminology

🏎️ Quick Start​

Prerequisites​

Before connecting, you'll need:

  1. Coalesce Catalog Account with API access
  2. API Token castor:xxx.. - see documentation here (do not share it, it is secret)
  3. MCP-Compatible AI Tool (Claude, ChatGPT, Cursor, Dust, …)

🔧 Setup Instructions​

Claude Desktop​

Claude Desktop does not support quick MCP setup for API token based authentification for now, we thus have to manually edit the configuration file.

Step 1: Create or edit your configuration file

Option A: direct access

  • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
  • Windows: %APPDATA%\Claude\claude_desktop_config.json
  • Linux: ~/.config/Claude/claude_desktop_config.json

Option B: via Claude Desktop settings:

  • Navigate to Claude Settings → Developer
  • "Edit Config"

Step 2: Add Coalesce MCP server

{
  "mcpServers": {
    "coalesce-catalog": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://api.castordoc.com/mcp/server",
        "--header",
        "Authorization: YOUR_API_TOKEN_HERE"
      ]
    }
  }
}

Step 3: Restart Claude Desktop

Step 4: Test the connection

Ask Claude: "Can you search for {your-data-topic} tables in Coalesce?"

Cursor​

Step 1: Open Cursor Settings

  • Navigate to Tools & MCP → New MCP server

Step 2: Configure the MCP server

Create or edit ~/.cursor/mcp.json:

{
  "mcpServers": {
    "Catalog": {
      "url": "https://api.castordoc.com/mcp/server",
      "headers": {
        "Authorization": "YOUR_API_TOKEN_HERE"
      }
    }  
  }
}

Step 3: Restart Cursor

Step 4: Verify in Cursor Agent

Open Cursor Agent and ask: "Find tables with {your-data-topic} data in Coalesce"

Dust​

  • Step 1: Access Dust workspace settings

    • Navigate to your Dust workspace
    • Go to Spaces → Tools → Add Tools

  • Step 2: Add new MCP connection

    Click Add Tools → Add MCP Server and fill in:

    • URL: https://api.castordoc.com/mcp/server
    • Authentication:
      • Type: Bearer Token
      • Bearer Token: YOUR_API_TOKEN_HERE

  • Step 3: Change name, tools stake settings if needed, Save changes

  • Step 4: Use in your Dust Agent

    • Go to Chat
    • Create or select an existing Agent → Edit Agent
    • Knowledge & Tools → Add Tools
    • Add the Catalog MCP
    • Test & Save

In any Dust conversation, your AI can now access Coalesce:

"@your-agent Search Catalog for {your-data-topic} tables tagged as {your-tag}"

🎖️ Authorizations​

The MCP server uses Public API Token authentication (OAuth not yet supported). There is no Role-Based Access Control (RBAC) - the server has full access to all data available through the API token, including all Catalog and API resources.

🔨 Available Tools​

The Coalesce MCP Server provides 8 specialized tools for exploring your data catalog:

Search Tools​

These tools use the same search engine as Coalesce Catalog's AI search, giving your AI assistant access to the same powerful search capabilities.

ToolPurposeKey Inputs
search_tablesFind data warehouse tablesquery (required), name, ownerName, tag, limit
search_dashboardsFind BI dashboardsquery (required), name, ownerName, tag, limit
search_columnsFind specific columnsquery (required), name, tag, limit
search_termsSearch business glossaryquery (required), name, ownerName, tag, limit

Catalog API Tools​

ToolPurposeInputs
get_company_contextRetrieve organizational context and documentationNone
get_tagsList tags used to categorize assetsids (optional), labelContains (optional)
get_usersGet all users and their profilesNone
get_teamsGet all teams and configurationsNone

😵‍💫 Troubleshooting​

A. Connection Issues​

Problem: "Failed to connect to MCP server"

Solutions:

  1. Verify API token is valid and not expired
  2. Check endpoint URL matches your Coalesce instance
  3. Test network connectivity to your instance

B. AI Not Using Tools​

Problem: AI doesn't leverage MCP tools effectively

Tips:

  • Be explicit: "Search Coalesce/Catalog MCP for…" (depending on how you named it in your settings)
  • Ask about specific assets: tables, columns, dashboards
  • Mention owners/teams by name
  • Reference tags when relevant

Support & Resources​

Documentation​

❔FAQ​

Q: Does this work with all AI assistants?

A: It works with any MCP-compatible client. We offer setup documentation for Claude, Cursor, and Dust. Others may work with custom configuration.

Q: What data does the AI assistant see?

A: The AI can search and read catalog metadata (table names, descriptions, owners, tags).

Q: What happens if my token expires?

A: The connection will fail with 401 Unauthorized. Generate a new token in your account settings and update your MCP configuration.

Last Updated: December 2025 | Version 0.0.1