This is the production environment!
Product image

Snowflake Text To SQL MCP Server for Enterprise

Snowflake Enterprise MCP Server by MCP4E turns natural language questions into instant, secure insights. It automatically converts plain English into SQL, executes queries in Snowflake with role-based access control, and delivers answers in real time. Empower non-technical users, accelerate decisions, boost productivity, and maintain enterprise-grade governance—all while maximizing the value of your AI and data investments.

Installation & Setup
Quick Start
# Install Directly
Download MCP4E Launcher. Run LICENSE_KEY=KEY_GOES_HERE ./mcp4e-launcher

# Or via Docker
docker run mcp4e/snowflake-enterprise-mcp-server
Enterprise Setup
  • SSO Configuration Guide
  • Audit Logging Setup
  • Security Policies
  • Compliance Documentation
View Full Documentation

Purchase

Choose your plan
Total: $500.00 / monthly + sales tax

Request a Demo

Interested in seeing this product in action? Schedule a personalized demo with our team.

Monday-Friday only

Product Description

Connect your AI projects with a Snowflake Database. Large Language Models can interact with tools that extend the capability to interact with resources outside of the LLM.

 

The Snowflake Enterprise MCP server connects the following concepts into one simple service:

  • Snowflake Analyst to translate plain text to a SQL query. Manage verified queries to help steer LLM questions towards the most appropriate query.
  • Snowflake Database - execute an RBAC controlled query on behalf of a user.
  • AWS S3 datastore - retain CSV files to support paginated outputs to the LLM or support friendly exports to your users leveraging AWS S3's secure pre-signed URLs for private content distribution of the generated SQL report output.

 

this service is an Enterprise offering and focuses on meeting enterprise needs of:

  • multiple users (per-user authorization)
  • per-user credentials management
  • integration with an OIDC provider (e.g., AWS Cognito).
  • Off-server object-based storage
  • per-query logging (the question and results are stored together on S3

 

Environment Variables

Environment variables can be set in one of two ways:

  1. Directly on the virtual machine or container running the mcp4e launcher process. E.g., KEY=VALUE LICENSE_KEY={somethingSecret} ./mcp4e-launcher.
  2. Via the management interface on mcp4e directly in your license configuration.

When to pick between which approach? We recommend setting SECURE values directly on the virtual machine. For other attributes that are not particularly secret, setting those values in the LICENSE configuration is an easy spot to manage configuration. Configuration is re-fetched whenever a new version of the MCP Server is found.

Variable Description Default
AWS_ACCESS_KEY_ID This is used for S3 operations. Recommended to provide this directly when running the application (none)
AWS_BUCKET some-s3-bucket to store the CSV reports and logs to (none)
AWS_REGION region of S3 bucket. Example: us-west-1 (none)
AWS_SECRET_ACCESS_KEY Secret for S3 operations. Recommended to provide this directly. (none)
AWS_SIGNING_DURATION_IN_MINUTES duration to allow downloading a report (presign url expiration) 1880
CORS_ALLOWED_ORIGIN Add if using JavaScript that calls the MCP Server directly to prevent CORS issues. (none)
SNOWFLAKE_ACCOUNT_ID Snowflake account ID. This is used as part of the URL when configuring the connection to snowflake YourAccountId
SNOWFLAKE_ANALYST_BASE_URL URL For Snowflake Analyst. E.G., "https://FMXXXX-HPBXXXX.snowflakecomputing.com" (none)
SNOWFLAKE_ANALYST_DEFAULT_ROLE Role to run the queries as in Snowflake Analyst (none)
SNOWFLAKE_ANALYST_SEMANTIC_MODEL_FILE This is the Semantic Model File that connects Analyst with the underlying data. This file needs to be carefully managed. Example: "@SOME_DATABASE.SOME_SCHEMA.SOME_STAGE/semantic_model_name.yaml" (none)
SNOWFLAKE_JDBC_ACCOUNT Your Snowflake Account ID. E.g., "FMXXXX-HPBXXXX" (none)
SNOWFLAKE_JDBC_DATABASE Database name goes here. (none)
SNOWFLAKE_JDBC_URL Database connection string. Example: "jdbc:snowflake://FMXXXX-HPBXXXX.snowflakecomputing.com" (none)
SNOWFLAKE_JDBC_USERNAME_PREFIX useful for setting a prefix when using RBAC. Example, "USER_ID_". This takes a cognito subject claim, converts hyphens to underscores and outputs a name like, "USER_ID_XXX_XXX_XXX". (none)
SNOWFLAKE_JDBC_WAREHOUSE Snowflake Warehouse to use. (none)
TOOL_SNOWFLAKE_DESCRIPTION Set this description to provide the LLM with more specific data-aware information. Perform data analysis against data stored in the Snowflake Data Warehouse. Accepts plain text, performs Text to SQL analysis using Snowflake Analyst, then executes the query on Snowflake via JDBC as the authenticated user.
TOOL_SNOWFLAKE_NAME Set this to override the Analyst (Text to SQL) and Query Execution tool name ask_snowflake_dw
TOOL_SQL_FETCH_PAGED_DESCRIPTION Describes how to fetch paginated results (JSON) (none)
TOOL_SQL_FETCH_PAGED_NAME describes how to fetch paginated results to LLM (none)
TOOL_SQL_QUERY_HISTORY_DESCRIPTION Describes the tool for fetching the user's query history (none)
TOOL_SQL_QUERY_HISTORY_NAME set the name for getting the user's query history (none)