Help & Documentation
Open Support Ticket

Setting Up Custom MCP Server for Snowflake (v7.2.0)

This article describes a third step of Connecting to Snowflake MCP Server. Set up and run the Custom Metric Insights MCP Server locally on the same machine where the application is running or remotely on a separate server or cloud platform.

If you want to use Snowflake Managed MCP Server, use this article instead.

Prerequisites:

Set Up the MCP Server

After choosing the platform:

1. Clone this repository:

git clone git@github.com:metricinsights/snowflake-cortex-mcp-server.git

2. Copy [config.example.py](https://github.com/metricinsights/snowflake-cortex-mcp-server/blob/main/config.example.py) to config.py folder and fill in the Snowflake account values.

Those values were gathered in Snowflake account.

3. Run the service via docker:

docker-compose up -d

Config Example

In this example three Agents are created: "CUSTOMER_SUPPORT", "INVENTORY", and "RETAIL_SALES". In your case those can be different in number and names.

_defaults = Config(
     account="TTXXXXX.us-east-2.aws",
     database="SNOWFLAKE_INTELLIGENCE",
     schema="AGENTS",
     token="<TOKEN>",
 )
 agent_configs = {
     "CUSTOMER_SUPPORT": Config(agent="CUSTOMER_SUPPORT", defaults=_defaults),
     "INVENTORY": Config(agent="INVENTORY", defaults=_defaults),
     "RETAIL_SALES": Config(agent="RETAIL_SALES", defaults=_defaults), }

NOTE: In the config example the Token is hardcoded into the server configuration. To prevent security risks, it is recommended to protect the server with a firewall.