How to Configure Windsurf Bigquery Sandbox

Published
Length
3 min read

Luis Felipe Dias

All posts

You set up a windsurf bigquery sandbox by linking your local shell or an MCP server to a Google Cloud test project. Your AI coding agent then uses your local cloud identity to read tables and run test queries. Teams often start by logging in with their own cloud profiles on their laptops. That quick setup gives the agent every right the user holds. If that role is too wide, a bad prompt can alter data or scan huge tables. How a windsurf bigquery sandbox works in practice On June 2, 2026, Cogn

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 sandbox by linking your local shell or an MCP server to a Google Cloud test project. Your AI coding agent then uses your local cloud identity to read tables and run test queries. Teams often start by logging in with their own cloud profiles on their laptops. That quick setup gives the agent every right the user holds. If that role is too wide, a bad prompt can alter data or scan huge tables.

How a windsurf bigquery sandbox works in practice

On June 2, 2026, Cognition renamed Windsurf to Devin Desktop. The app keeps the same IDE and features under the new name. Inside the editor, the agent can run terminal commands on its own with user approval. It opens a dedicated shell on your machine, using zsh on macOS or bash on Linux. Because this shell loads your normal profile, the agent inherits your local aliases and variables.

This shell setup is the first path to your data warehouse. When you log in to Google Cloud on your laptop, your shell stores your active token. The agent can then run command-line queries against your test datasets using your profile. The second path uses the Model Context Protocol. The legacy Cascade agent connects to MCP servers that you define in your local settings. An MCP server for BigQuery takes tool calls from the agent and sends SQL to Google Cloud. Both paths rely on the cloud keys or tokens stored on the developer desktop.

What native cloud roles solve and what they miss

Google Cloud offers a sandbox mode and project roles to limit what a user can do. You can place test tables in a separate project and grant read-only access to the team. This native setup decides who can log in and which datasets they can reach. It also helps cap storage costs and blocks access to billing settings in test spaces.

Still, static cloud roles do not check individual SQL queries while they run. Once a user or service key logs in, the database trusts every command that role allows. If a developer connects their shell to a shared project, the agent gets that same broad reach. A prompt mistake can trigger a full table scan or pull sensitive rows into the editor context. Test tables often hold copies of real customer emails, phone numbers, card numbers, or national IDs. Native roles grant or deny whole tables, so they do not mask those fields in query results on the fly.

BigQuery IAM + AI Sandbox Environments: architecture patterns and best practices

Continue reading? Get the full guide.

Free. No spam. Unsubscribe anytime.

Why local agent credentials create blind spots

Storing database keys or cloud tokens on every laptop creates extra work for security teams. Each MCP server needs a key or an active login session to talk to the cloud. When ten engineers set up the editor, ten machines hold live credentials. If one token leaks, someone can use it outside the editor without warning.

Cloud audit logs also lose clarity when agents share a developer identity. Every query from the agent looks like the human typed it by hand. Teams cannot tell which SQL came from a prompt and which came from a manual check. Rotating shared service keys is also hard because updating them breaks local editor setups across the team.

Check database and API queries on the wire

Identity tools, roles, and network rules decide who opens a session. Setup alone cannot check what each query does or mask sensitive values in a response once a session starts. When workflows touch Postgres, MySQL, SQL Server, MongoDB, ClickHouse, Kubernetes APIs, SSH, or internal HTTP APIs, hoop.dev sits inline in the wire protocol beside those services as an open-source proxy, a sidecar for your Postgres or API. On every query an AI agent or engineer runs against production, hoop.dev decides to pass, mask, block, or hold it for human review.

Your login details and client tools stay as they are while hoop.dev guardrails refuse bad deletes or drops before they run, and hoop.dev data masking rewrites matched sensitive values in responses in memory. Furthermore, hoop.dev holds SOC 2 Type II, is MIT licensed, and writes every query and decision to an audit trail that generates evidence for HIPAA, GDPR, and PCI DSS. See the source on GitHub.

Frequently asked questions

Is Windsurf the same tool as Devin Desktop?

Yes, Cognition announced on June 2, 2026, that Windsurf is now called Devin Desktop. It keeps the same IDE, terminal runner, and Cascade agent features.

How does the agent reach BigQuery from the editor?

The agent either runs cloud commands in its built-in shell or calls a configured MCP server. Both methods use the cloud credentials available on your local machine.

Why are read-only cloud roles not enough for AI agents?

Read-only roles still let an agent pull whole tables into the model prompt. They do not block costly scans or rewrite emails and card numbers in results.

The query runs. The emails stay hidden.