MCP Environment Setup

Install dependencies and configure database parameters for the Arivu MCP Server.

Installation

The Arivu MCP server requires the mcp optional dependency map.

terminal

pip install "arivu-ai[mcp]"

The MCP server runs globally within your Python environment. For best results, install it within the same virtual environment where you intend to launch Claude Desktop or Cursor.

Global Configuration

The Arivu MCP Server does not take arguments programmatically; instead, it binds to Environment Variables securely passed in by the parent MCP client.

Core Variables

.env

# Common configuration applied across all dialects
ARIVU_DB_DIALECT="postgresql" # postgresql, mysql, sqlite, databricks, snowflake
ARIVU_DB_MODE="user"          # Safety guardrail (admin bypasses approval restrictions)

Dialect-Specific Configurations

PostgreSQL / MySQL

Standard RDBMS parameters:

ARIVU_DB_HOST="localhost"
ARIVU_DB_PORT="5432"
ARIVU_DB_USER="postgres"
ARIVU_DB_PASSWORD="supersecretpassword"
ARIVU_DB_NAME="analytics"

Snowflake

Requires warehouse routing context:

ARIVU_SNOWFLAKE_ACCOUNT="xy12345.us-east-1"
ARIVU_SNOWFLAKE_WAREHOUSE="COMPUTE_WH"
ARIVU_DB_USER="admin"
ARIVU_DB_PASSWORD="mypassword"
ARIVU_DB_NAME="SNOW_DATA"

Databricks

Requires Databricks catalog tracking:

ARIVU_DB_HOST="adb-123.azuredatabricks.net"
ARIVU_DATABRICKS_HTTP_PATH="/sql/1.0/endpoints/abc123"
ARIVU_DATABRICKS_TOKEN="dapi..."
ARIVU_DATABRICKS_CATALOG="main"

Server Boot Modes

Arivu supports both standard execution layers as outlined in the MCP specification.

1. Stdio Transport (Local)

   Standard mechanism for desktop integrations like Cursor. It reads inputs on stdin and writes on stdout.

   python -m arivu.mcp
   
2. SSE HTTP Transport (Remote)

   Used if you are hosting the MCP Server exclusively on a remote VPC droplet.

   python -m arivu.mcp --transport http --port 8080