BigQuery MCP Server

No. Google offers a Google-managed remote BigQuery MCP server and the open-source MCP Toolbox for Databases. InsightfulPipe is an independent, hosted server that connects to the BigQuery API with your service account. InsightfulPipe is not affiliated with Google.

Google's remote server is enabled with the BigQuery API, uses OAuth with IAM and needs the MCP Tool User role, includes a read-only SQL tool, and limits queries to three minutes and 3,000 rows. MCP Toolbox for Databases runs on infrastructure you manage. InsightfulPipe is hosted, connects with a service account key, returns up to 10,000 rows per query, and adds per-project action controls and team roles.

Any AI client that supports remote MCP servers. The setup guide on this page covers Claude, ChatGPT, Claude Code, Cursor, Gemini CLI and Codex. You can also work with the same connected accounts from the terminal with the InsightfulPipe CLI.

Yes. In Claude, add https://bigquery.insightfulmcp.com/ as a custom connector and authorize your InsightfulPipe account. You can also add InsightfulPipe from the Claude Directory, which connects all of your linked accounts at once.

Yes. Add the server URL as a custom connector in ChatGPT Developer mode, or add InsightfulPipe from the ChatGPT app directory to connect all of your linked accounts. Developer mode availability depends on your ChatGPT plan.

Anything the service account can read in the connected project: the list of datasets, tables and views, table schemas and metadata, and query results from Standard SQL. The server has five actions: list datasets, list tables, get table schema, get table metadata and run query.

Only if the service account's IAM roles allow it. A single SELECT runs as a read. Every other statement, including DML, DDL and multiple statements in one request, is treated as a write that runs only through the write tool and only for owners, admins and operators.

Give the service account read roles such as BigQuery Data Viewer and BigQuery Job User, so BigQuery itself rejects writes. You can also turn off the run query action for the project in InsightfulPipe, which leaves only schema and metadata actions.

Yes. Each connection is one Google Cloud project with its own service account key. Connect as many projects as you need, and each counts as one connection on your plan.

You upload or paste a service account JSON key. InsightfulPipe checks the key with Google, stores it encrypted and uses it to call the BigQuery API for that project. Your AI client signs in to the MCP server with your InsightfulPipe account and never sees the key.

Queries return at most 10,000 rows, and a query that runs longer than its wait time returns a job reference without rows. There is no bytes-billed cap, and Google bills queries to your project as usual. There are no separate actions for creating datasets or listing jobs, and new connections use a service account rather than a Google sign-in.

The BigQuery MCP server is included in every InsightfulPipe plan, together with all other MCP servers and the CLI. Plans start at $29.99/month and include a trial. See insightfulpipe.com/pricing for current plans.