How to Configure Goose Snowflake Approval

Published
Length
3 min read

Luis Felipe Dias

All posts

You set up a goose snowflake approval flow by requiring human review before queries run. This setup stops the AI agent from changing data or reading tables without a check. Many teams start by giving the local agent a standard database role. If that role is too broad, the agent can read or drop tables by mistake. A clear review step keeps your data safe while the agent works. How a native goose snowflake approval flow works Goose runs on a developer machine and uses tools to talk to external

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 goose snowflake approval flow by requiring human review before queries run. This setup stops the AI agent from changing data or reading tables without a check. Many teams start by giving the local agent a standard database role. If that role is too broad, the agent can read or drop tables by mistake. A clear review step keeps your data safe while the agent works.

How a native goose snowflake approval flow works

Goose runs on a developer machine and uses tools to talk to external systems. When it needs warehouse data, it drafts a query and prepares to run it. You can set up the local agent so it pauses before it runs tools. The agent shows the planned query to the user in the terminal window. The user reads the text and says yes or no to the action.

On the database side, admins set up specific roles for the agent session. You log in the session using a dedicated user profile with limited warehouse permissions. The database checks if this role has the right to read or write tables. Still, the database itself does not pause a valid query to ask a human. The native review step happens only on the local machine of the developer.

What native roles solve and what they miss

Database roles do a good job of blocking tables the user should never touch. If a role lacks write permissions, the agent cannot drop a table by mistake. Local terminal prompts also help developers catch obvious syntax mistakes before a query runs. Together, these two steps stop basic errors during daily data tasks.

Still, a big gap remains when you rely on local prompts and static roles. Database roles decide who logs in, not what each query does in flight. A developer might approve a fast terminal prompt without reading the full query text. Also, team leads cannot review what the agent runs on a personal laptop. If a table holds emails, phone numbers, or card numbers, the query returns them. Static roles do not rewrite matched fields in results when the agent reads rows.

Approval Chains & Escalation + Snowflake Access Control: architecture patterns and best practices

Continue reading? Get the full guide.

Free. No spam. Unsubscribe anytime.

Building a team review workflow for agent queries

To fix this, teams move the approval step off the local developer laptop. Instead of a quick terminal prompt, risky queries should route to a central peer review. When the agent drafts a command that changes data, a teammate checks it first. The reviewer reads the exact statement in a chat app and clicks approve.

This shared flow gives security teams a clean log of every approved action. You see who asked for the query, who checked it, and when it ran. You can also set rules to mask emails or national IDs in query results. Masking rewrites matched fields in results so raw sensitive values stay protected during runs. This keeps agent sessions fast for safe reads while guarding high-risk warehouse actions.

Govern statements and responses across databases and APIs

Good config and scoped roles are needed setup. They do not check single queries in flight or mask sensitive fields in returned rows. Across Postgres, MySQL, SQL Server, MongoDB, ClickHouse, Kubernetes APIs, SSH, and internal HTTP endpoints, hoop.dev runs beside each supported service as an open-source proxy, a sidecar for your database or API.

First, hoop.dev guardrails block banned queries before they run. Then hoop.dev data masking rewrites matched sensitive values in responses in memory. The audit trail records every statement and verdict per resource. Furthermore, hoop.dev holds SOC 2 Type II and is open source on GitHub.

Frequently asked questions

Can Snowflake pause a single query for human approval on its own?

No, the database checks role permissions right away and runs valid queries at once. To pause a single statement for human review, you need an external workflow step.

Why is a local Goose prompt not enough for production data?

A local prompt only asks the person who started the agent session to click yes. It does not require a second teammate to check risky changes or log the review.

How do we protect sensitive fields when the agent reads tables?

You can apply data masking rules that inspect results before they reach the local agent. Masking rewrites matched fields like emails, phone numbers, and card numbers in the output.

The query runs. The emails stay hidden.