Getting started with our MCP server

dScribe also exposes an MCP (Model Context Protocol) server, letting AI assistants and agent workflows query your catalog metadata directly. It's reachable over Streamable HTTP at:

https://app.dscribe.cloud/catalog/api/mcp
https://<tenant>.dscribe.cloud/catalog/api/mcp for tenant-scoped environments

Currently it exposes a single tool, searchGraph, which takes a natural-language question and returns relevant catalog assets and metadata — descriptions, custom properties, contracts, SLAs, and data quality checks — making it useful as a first step for grounding an AI workflow in your catalog before acting on that context (think chat with your data).

Authenticating a chatbot

As described in our API docs, you have two methods to authenticate with our API. The same goes for our MCP server.

Via API key

This is an example in Claude Desktop.

Open claude_desktop_config.json (on macOS: ~/Library/Application Support/Claude/claude_desktop_config.json; on Windows: %APPDATA%\Claude\claude_desktop_config.json). Replace <your-dscribe-api-key> with a key generated from dScribe (Settings → API keys — it will start with fd__), and swap the host for your tenant-scoped URL (https://customer-api.<tenant>.dscribe.cloud/api/mcp) if applicable. Save the file and fully restart Claude Desktop for it to pick up the new server — once connected, Claude will have access to the searchGraph tool and can use it to ground answers in your catalog's assets, contracts, and data quality metadata.

{
  "mcpServers": {
    "dscribe": {
      "url": "https://app.dscribe.cloud/catalog/api/mcp",
      "headers": {
        "x-dscribe-api-key": "<your-dscribe-api-key>"
      }
    }
  }
}

The same goes for other chatbots as well. Find the place where you can add custom headers and pass our api key and url.

Via OAuth Client

You can find a guide here on how to add our MCP server to Databricks Genie (but the same setup goes for other chatbots as well).

You need the following information and paste it in the appropriate fields:

Host	https://app.dscribe.cloud
Port	443
Client ID	<ID from Oauth clients page>
Client Secret	<Secret from Oauth clients page>
OAuth scope	*
Token URL	https://app.dscribe.cloud/catalog/api/api-token
Base Path	/catalog/api/mcp

A token will be fetched under the hood and your chatbot will be authenticated with dScribe to use your catalog data.