Model Context Protocol moderate risk

Snowflake

Snowflake MCP server enables data interaction and analysis via SQL queries and schema exploration, with configurable write access and insight aggregation.

Connections & Capabilities

Connects To

GitHubTwitter

Capabilities

readwriteexec

Quickstart

Install

npx -y @smithery/cli install mcp_snowflake_server --client claude

Exposed MCP Tools (8)

safe

read_query

Executes SELECT queries to retrieve data from the Snowflake database.

Read-only operation, no data modification.

high

write_query

Executes INSERT, UPDATE, or DELETE queries to modify data in the Snowflake database.

Allows modification and deletion of data.

high

create_table

Creates new tables within the Snowflake database.

Allows structural changes to the database schema.

safe

list_databases

Lists all databases available in the Snowflake instance.

Read-only operation, no data modification.

safe

list_schemas

Lists all schemas within a specified database.

Read-only operation, no data modification.

safe

list_tables

Lists all tables within a specified database and schema.

Read-only operation, no data modification.

safe

describe_table

Retrieves column information for a specified table.

Read-only operation, no data modification.

moderate

append_insight

Adds a new data insight to the insights memo.

Non-destructive write operation, but could be used to inject misleading information.

Safety Assessment

The Snowflake MCP server is relatively safe in its default read-only configuration. Enabling write access significantly increases the risk, requiring careful management of permissions and input validation to prevent data corruption or unauthorized modifications. Use of connection files is recommended to avoid hardcoding credentials.

Read-only access by default
Option to disable specific tools
Logging capabilities for auditing
Connection parameters can be managed via TOML files for better security
Exclusion patterns for databases, schemas, or tables

Write access can be enabled, posing a risk of data modification or deletion
Direct SQL execution can lead to injection vulnerabilities if queries are not properly sanitized
Sensitive credentials (passwords, private keys) need to be securely managed
Lack of built-in rate limiting could lead to abuse
No explicit sandboxing of SQL execution