How to Configure Windsurf Bigquery MCP Server

Published
Length
3 min read

Luis Felipe Dias

All posts

You set up a windsurf bigquery mcp server by registering the server in your editor settings. Then you sign in to Google Cloud on your local machine. Cognition renamed Windsurf to Devin Desktop, though the editor keeps the same features. Its Cascade agent reads your local MCP settings file to start the tool process. Teams often assign broad cloud roles to get this local setup working fast. Broad roles let the agent read or change every table in your project. How the windsurf bigquery mcp server

Your coding agents run with your access.

Claude Code reaches prod with a developer's credentials, and approval prompts get clicked through. hoop.dev checks every query, masks sensitive data, and holds risky changes for a human.

You set up a windsurf bigquery mcp server by registering the server in your editor settings. Then you sign in to Google Cloud on your local machine. Cognition renamed Windsurf to Devin Desktop, though the editor keeps the same features. Its Cascade agent reads your local MCP settings file to start the tool process. Teams often assign broad cloud roles to get this local setup working fast. Broad roles let the agent read or change every table in your project.

How the windsurf bigquery mcp server setup works

Windsurf, now called Devin Desktop, uses the Cascade agent to run local tools. The agent reads your local MCP settings to find external servers. When you start a session, the editor launches the BigQuery MCP server process. This process runs on your workstation under your local user account.

The agent can also run terminal commands on its own with user permission. It uses a dedicated shell on macOS or Linux. That shell loads your user profile, including your shell aliases and variables. Because of this, the MCP server inherits your local cloud credentials and keys. You either log in through the cloud CLI or pass a service account key.

Once the server starts, the Cascade agent can list datasets and inspect table schemas. The agent writes SQL queries based on your prompts and sends them to BigQuery. BigQuery checks the cloud role tied to your local identity. If the role allows the action, BigQuery runs the query and returns the rows. The MCP server then passes those data rows back into the agent context.

What native cloud roles solve and what they leave open

Google Cloud IAM roles control who can log in and reach a project. You can give a user or service account a read-only role on one dataset. You can also block access to whole projects. This setup stops unauthorized users from reaching your data warehouse.

Still, cloud IAM checks identity at the start of a request, not query intent. Once a role grants read access, the agent can query every column in that dataset by default. A single prompt can pull emails, phone numbers, card numbers, or national IDs. Those raw values flow right into the model context and external AI logs. Standard cloud roles cannot rewrite matched fields in query results on the fly.

BigQuery IAM + Kubernetes API Server Access: architecture patterns and best practices

Continue reading? Get the full guide.

Free. No spam. Unsubscribe anytime.

Native roles also struggle with shared developer credentials and local keys. Engineers often store static service account keys on their laptops for MCP tools. Those keys sit in plain text files and get loaded into the agent shell. If the agent runs a bad command, it acts with the full power of that key. Static roles cannot pause a dangerous query or require peer approval for one statement.

Why local shell inheritance increases data risk

The Devin Desktop terminal loads your standard shell profile when it runs commands. Any secret or token in your shell profile becomes available to the agent. If you hold admin roles for daily work, the agent inherits those exact rights. Even if you limit the MCP server, the agent can run direct shell commands.

Teams need a way to inspect every query before it hits the warehouse. They also need to mask sensitive fields like emails and card numbers in results. Separating developer identity from raw database credentials keeps local setups safe.

Add runtime guardrails and masking beside your databases and APIs

Login rules, role grants, and CI checks govern who connects to your stack. Yet any allowed script, engineer, or AI agent still runs commands with the full scope of its role. For workloads on Postgres, MySQL, SQL Server, MongoDB, ClickHouse, Kubernetes, SSH, or internal HTTP services, hoop.dev sits inline in the protocol beside the service as an open-source proxy, a sidecar for your database or API. On every production query an agent or human runs, hoop.dev decides whether to pass, mask, block, or escalate it for review.

Before a query runs, hoop.dev guardrails check the statement against your rules, an AI Analyzer you configure can score intent or hold risky calls for human approval, and hoop.dev data masking rewrites sensitive fields in responses in memory. Explore the MIT-licensed project on GitHub.

Frequently asked questions

Did Windsurf change its name to Devin Desktop?

Yes, Cognition renamed Windsurf to Devin Desktop on June 2, 2026. The editor keeps the same features, and the Cascade agent still uses MCP servers.

How does the Cascade agent authenticate to BigQuery?

The local MCP server process uses the cloud credentials on your machine. It reads your active cloud CLI login or a service account key file.

Can Google Cloud IAM hide sensitive columns from the AI agent?

Standard project and dataset roles grant access to whole tables at once. Unless you build complex policy tags for every column, raw data reaches the agent.

The query runs. The emails stay hidden.