Igloo

Igloo MCP is a lightweight MCP server that connects your AI assistant to Snowflake with built-in safety, caching, and auditing. It enables querying databases, building catalogs, and creating living reports through natural language, while protecting you from unsafe queries and keeping a detailed audit trail.

How to use

You interact with Igloo MCP through your preferred MCP client (Cursor, Claude, or other compatible runners). Start by ensuring your Snowflake connection is configured, then run the Igloo MCP server profile to enable query execution, catalog access, and living reports. You can run queries with guardrails to prevent dangerous actions, build offline catalogs for fast searches, and evolve auditable living reports with full history.

How to install

Prerequisites: ensure you have Python and a compatible MCP client environment ready. You will install the Igloo MCP package, configure Snowflake access, and then wire Igloo MCP into your MCP clients.

# Install igloo-mcp
pip install igloo-mcp

# Configure Snowflake connection (uses Snowflake CLI)
snow connection add --name quickstart --account <account> --user <user> --authenticator externalbrowser --warehouse <warehouse>

Additional setup snippets you will use

{
  "mcpServers": {
    "igloo-mcp": {
      "command": "igloo-mcp",
      "args": ["--profile", "quickstart"]
    }
  }
}

Configure clients to use Igloo MCP

Cursor and Claude Code can both connect to Igloo MCP to run queries, build catalogs, and manage Living Reports. Use the provided profile and start the MCP server with the profile you configured in Snowflake. Restart the clients after you add or update the MCP server configuration.

Security and auditing highlights

Igloo MCP includes built-in safety features that block dangerous DDL/DML by default, auto-cancel slow queries, and log every execution. Every execute_query call is recorded in query history, and Living Reports maintain complete audit trails for modifications via evolve_report.

Use test_connection to validate authentication before running queries and rely on progressive disclosure and smart result caching to reduce token usage while keeping results auditable.

Living Reports and data access

Create auditable, evolving business reports with create_report, modify them safely using evolve_report, attach charts to insights, and export to HTML, PDF, or Markdown via render_report. You can search, retrieve, and cite reports to maintain a complete line of sight into your analytics work.

Notes for quick start usage

If you are using Cursor, restart Cursor after adding Igloo MCP configuration and then run a quick test like asking to preview a table. For Claude Code, restart Claude Code after adding Igloo MCP and then ask for a database schema or a list of available tables.

Available tools

execute_query

Run SQL queries with guardrails, timeouts, and auto-insights to ensure safe and efficient data access.

build_catalog

Export Snowflake metadata for offline search and quick discovery of objects.

search_catalog

Find tables and columns without querying Snowflake directly to speed up explorations.

build_dependency_graph

Visualize table lineage and dependencies to understand data flows.

create_report

Initialize auditable, JSON-backed Living Reports for ongoing analysis.

evolve_report

Safely modify Living Reports with an audit trail and guided edits.

evolve_report_batch

Apply multiple report changes atomically with a single operation.

render_report

Export reports to HTML, PDF, or Markdown via Quarto for sharing.

get_report

Read reports with various progressive disclosure modes to control detail.

search_report

Find reports by title or tags to locate prior work quickly.

search_citations

Search citations by source type or provider to reference data sources.

get_report_schema

Discover valid report structures and section templates at runtime.

test_connection

Validate Snowflake authentication before using Igloo MCP.

health_check

Monitor server, profile, and catalog status to ensure reliability.