At a glance
- Google’s remote BigQuery MCP server lives at bigquery.googleapis.com/mcp and is enabled with the BigQuery API.
- Google recommends a separate identity for agents that use MCP tools, so access can be controlled and monitored.
- Rig connects with a read-only service account, builds Rig Map over your datasets and gives every agent its own role, token, daily cap and audit trail.
What is BigQuery MCP?
Google runs a remote MCP server for BigQuery at https://bigquery.googleapis.com/mcp. Its tools list datasets and tables, return dataset and table information, and run SQL: execute_sql, and execute_sql_readonly for read-only work. Google’s page says execute_sql is the only tool that is not read-only, and suggests a deny policy if you want to block it. Both SQL tools have a three-minute processing limit and return up to 3,000 rows.
How does BigQuery MCP handle identity?
Authentication is OAuth 2.0 with IAM. Callers need the MCP Tool User, BigQuery Job User and BigQuery Data Viewer roles. Google recommends creating a separate identity for agents that use MCP tools, so that access can be controlled and monitored. We agree, and Gateway is how Rig does that for agents connected through Rig: each one gets an identity, a role and a log of its own.
Why give each BigQuery agent a role and a log?
BigQuery bills by the work a query does, and an agent can run a lot of queries. A shared service account also makes the audit log hard to read, because every agent looks the same. With a role per agent, a sales agent reads the sales datasets and nothing else, a daily cap limits how many calls it makes, and the log shows which agent asked what.
How Rig does it on BigQuery
Rig connects to BigQuery with a read-only service account, reads your datasets through INFORMATION_SCHEMA, and builds Rig Map over your tables, joins and metrics. Queries run as standard BigQuery SQL in your own project. Gateway sets what each person and agent can reach: schemas, columns, masked fields and rows. An agent connects with its own token and daily call cap, looks up the certified metric before it writes SQL, and each call is logged with its name and tool.
Who runs this on BigQuery?
Novakid runs BigQuery. Before opening the warehouse to sales, Novakid narrowed what the AI could see to well-described tables it trusted. Its sales managers and sales ops now investigate their own numbers in Claude, and the team has made 22,412 requests since March, with more than 4,000 a month since May.
Common questions
What is BigQuery MCP?
It is Google’s remote MCP server for BigQuery, at bigquery.googleapis.com/mcp, with tools to list datasets and tables, read their metadata and run SQL.
How do I give agents BigQuery access with a named role and a log?
Create each agent in Rig with its own role, token and daily call cap, and connect it to Rig’s MCP URL. Rig queries BigQuery with a read-only service account and logs every call under the agent’s name.
Can I use Google’s BigQuery MCP server and Rig together?
Yes. Some teams use Google’s server for engineering work in BigQuery and Rig for business teams and agents that need certified metrics and per-agent roles.
How do I expose governed business metrics to Claude and Cursor?
Certify the metrics once in Rig Map, then connect Claude or Cursor to Rig’s MCP server. A person signs in with OAuth and gets their own role. An agent gets its own role and token. Either way, the tool looks up the certified metric before it writes SQL, sees only the schemas, columns and rows the role allows, and every call lands in the audit log.
Sources and product details
Vendor information checked on 20 November 2026. Confirm current editions, licence terms and prices with the vendor before buying.