Dreamdata connects its data warehouse to AI agents through MCP

Dreamdata published an engineering blog post on February 18, 2026 explaining how customers with its Data Warehouse add-on can connect that data to AI agents through the Model Context Protocol. The company says the warehouse, which customers can reach through Google BigQuery, Amazon S3, Snowflake and Microsoft Azure Storage, is implemented on BigQuery. That lets Dreamdata use Google's managed MCP server for BigQuery to connect agents such as Claude, ChatGPT and Gemini. The post walks through enabling the MCP API in Google Cloud, creating an OAuth client, granting IAM roles and adding a custom connector in Claude. Dreamdata describes the setup as text-to-SQL and notes model outputs must be checked.

The details

• Requires the Data Warehouse add-on with the BigQuery warehouse enabled.
• The MCP endpoint is Google's hosted server at bigquery.googleapis.com/mcp, enabled with gcloud commands.
• Users need the MCP Tool User, BigQuery Job User and BigQuery Data Viewer IAM roles.
• Agents can access the same warehouse data as the authenticating user.
• Available tools are limited to those Google's hosted BigQuery MCP server exposes.
• Dreamdata says a semantic layer would generally produce better results than this text-to-SQL setup.

Context

Earlier Dreamdata coverage centers on November 2025 releases, including Activation for account-based campaigns, the Audience Reach Report and CRM Syncs for HubSpot and Salesforce. This post preceded the Dreamdata AI launch on August 31, 2026, which added a native MCP server and an Analytics Agent. In between, Dreamdata published LinkedIn Ads benchmarks in March and launched Analytics Hub in June.

Not disclosed

• Price of the Data Warehouse add-on
• Whether Dreamdata supports the setup or it is self-serve only
• Equivalent setup steps for ChatGPT or Gemini
• Google Cloud costs incurred by queries