Skip to main content
DataDoe HubHub
(opens in a new tab)
ChatGPT (opens in a new tab)
Claude (opens in a new tab)

Using MCP Toolbox with BigQuery

MCP Toolbox for Databases lets an AI assistant explore and query your DataDoe linked dataset in BigQuery.

Before you start

  • Connect to BigQuery and subscribe to the DataDoe listing in your Google Cloud project
  • Set up Google Cloud Application Default Credentials for an account that can query the linked dataset
  • Install Node.js so your AI assistant can run npx

Add MCP Toolbox to your AI assistant

Add this server to your assistant's MCP configuration file. The surrounding configuration may differ by assistant.

json
1{
2    "mcpServers": {
3        "toolbox-bigquery": {
4            "command": "npx",
5            "args": ["-y", "@toolbox-sdk/server", "--prebuilt=bigquery", "--stdio"],
6            "env": {
7                "BIGQUERY_PROJECT": "<YOUR_GOOGLE_CLOUD_PROJECT_ID>",
8                "BIGQUERY_LOCATION": "us-east4",
9                "BIGQUERY_READONLY": "true"
10            }
11        }
12    }
13}

Replace <YOUR_GOOGLE_CLOUD_PROJECT_ID> with the project where you created the linked dataset. Set BIGQUERY_LOCATION to the linked dataset's region if you chose a region other than us-east4. Read-only mode prevents Toolbox from running write queries. See the BigQuery prebuilt configuration (opens in a new tab) for other options.

Reload your AI assistant

Restart or reload your assistant so it starts MCP Toolbox. For example:

  • In Cursor or VS Code, reload the window
  • In Claude Desktop, quit and reopen the app
  • In Claude Code, start a new session or check the server with /mcp

Give the assistant data guidance

Download the DataDoe AGENTS.md file and add it to your assistant's project instructions. It tells the assistant to use your linked dataset, check the current data scheme (opens in a new tab), and run read-only queries.

Your linked dataset uses the name you chose when subscribing. It may not end in _integrated or _raw. The available tables depend on your DataDoe connections and whether you shared Integrated or Raw data.

Check the connection

Ask your assistant to list datasets, then name the linked dataset you created:

List the datasets in my Google Cloud project. Then list the tables in my DataDoe linked dataset.

MCP Toolbox provides list_dataset_ids, list_table_ids, get_table_info, and execute_sql for this workflow. Ask the assistant to inspect a table before writing a query, and to keep query results small.

If the dataset is missing, check the subscription and Google account. If the server fails to start or cannot authenticate, see the MCP Toolbox documentation (opens in a new tab).

Related guides