Igloo MCP

# Getting Started with igloo-mcp > **Quick Start**: Set up your Snowflake profile → Install igloo-mcp → Start using with your AI assistant ## How It Works Your LLM calls MCP tools (execute_query, build_catalog, evolve_report, etc.) through igloo-mcp, which uses your Snowflake CLI profile for authentication. Built-in guardrails block risky SQL operations, while timeouts and cancellation keep runs responsive. Every query logs to JSONL for auditing. ## Prerequisites **Required**: 1. **Python 3.12+** with `uv` or pip 2. **Snowflake CLI** (bundled with igloo-mcp) 3. **Snowflake account** with appropriate permissions (USAGE on warehouse/database/schema, SELECT on INFORMATION_SCHEMA) 4. **AI Assistant** that supports MCP (Cursor, Claude Code, etc.) ## Installation & Setup For complete installation instructions, see the **[Installation Guide](installation.md)**. **Quick install**: ```bash uv pip install igloo-mcp ``` **Quick Snowflake profile** (SSO recommended): ```bash snow connection add --connection-name my-profile --account <account>.<region> --user <username> --warehouse COMPUTE_WH --authenticator externalbrowser ``` **Quick MCP config** (Cursor example - see [Installation Guide](installation.md) for other clients): ```json { "mcpServers": { "igloo-mcp": { "command": "igloo-mcp", "args": ["--profile", "my-profile"] } } } ``` > **Note**: For detailed setup including all authentication methods, multiple profiles, and troubleshooting, see [Installation Guide](installation.md). ## Verify Your Setup Once configured, verify igloo-mcp works correctly through your AI assistant: ### Health Check (Recommended First Step) ``` "Run a health check" → Uses: health_check → Validates: Snowflake connectivity, profile configuration, catalog availability ``` This verifies your server, profile authentication, and catalog status in one call—ensuring everything works before running queries. ### Example Prompts ``` "Test my Snowflake connection" → Uses: test_connection "Show me the first 10 rows from CUSTOMERS table" → Uses: execute_query with auto-insights "Build a catalog for MY_DATABASE" → Uses: build_catalog "Find all tables with 'user' in the name" → Uses: search_catalog "Create a dependency graph for the ANALYTICS schema" → Uses: build_dependency_graph ``` ## Available MCP Tools Igloo MCP provides **14 focused tools** organized by workflow. See [API Tools Index](api/TOOLS_INDEX.md) for complete workflow documentation. ### 🔍 Data Discovery - `test_connection`, `execute_query`, `build_catalog`, `search_catalog`, `build_dependency_graph` ### 📊 Living Reports - `create_report`, `search_report`, `get_report`, `get_report_schema`, `evolve_report`, `evolve_report_batch`, `render_report` ### 🏥 Health & Diagnostics - `health_check`, `get_catalog_summary` **View complete workflows**: [API Tools Index](api/TOOLS_INDEX.md) --- ## Token-Efficient Report Workflows Use progressive disclosure to minimize token usage in multi-turn Living Reports workflows: ```python # Efficient workflow example # 1. Find reports (minimal fields) reports = search_report( title="Q1 Sales", fields=["report_id", "title"] ) # 2. Inspect with summary mode summary = get_report( report_selector=reports["reports"][0]["report_id"], mode="summary" ) # 3. Discover valid schemas schema = get_report_schema( schema_type="proposed_changes", format="examples" ) # 4. Evolve with minimal response result = evolve_report( report_selector=reports["reports"][0]["report_id"], instruction="Add revenue insight", proposed_changes={...}, # Based on schema response_detail="minimal" ) ``` **Key Features**: - **Progressive disclosure**: `get_report` with 4 modes (summary/sections/insights/full) - **Schema discovery**: `get_report_schema` for runtime API introspection - **Field filtering**: `search_report` with selective field retrieval - **Response control**: `evolve_report` with configurable verbosity - **Preview sizing**: `render_report` with adjustable preview truncation See [Living Reports User Guide](living-reports/user-guide.md) for complete workflows. ## Advanced Configuration For configuration options including multiple profiles, environment variables, and custom storage paths, see the [Configuration Guide](configuration.md). ## Troubleshooting ### MCP Server Won't Start **Solution**: 1. Verify Snowflake profile: `snow connection list` 2. Test connection: `snow sql -q "SELECT 1" --connection my-profile` 3. Check MCP configuration in your AI assistant settings 4. See [Installation Guide](installation.md#troubleshooting) for detailed troubleshooting ### Authentication Errors **Solution**: See [Authentication Guide](authentication.md) for all auth methods and troubleshooting. ### Tool Not Found **Solution**: Restart your MCP client completely after configuration changes. ## Next Steps - 🎯 [Installation Guide](installation.md) - **Complete setup documentation** - 🔧 [API Reference](api/README.md) - All MCP tools documentation - 📊 [Living Reports User Guide](living-reports/user-guide.md) - Report workflows - 💡 [Examples](examples/catalog-examples.md) - Real-world usage examples ## See Also - [Installation Guide](installation.md) - Detailed installation and profile setup - [MCP Integration Guide](mcp-integration.md) - General MCP client configuration - [Authentication Guide](authentication.md) - Authentication options and troubleshooting - [Features Overview](features_overview.md) - Complete feature list ## Administrative CLI Tools igloo-mcp includes administrative CLI tools for power users and system administrators. The primary interface for development work is through MCP tools in your AI assistant. See the [Administrative Operations](docs/index.md#administrative-operations) section for CLI usage details. --- *Questions? Check our [GitHub Discussions](https://github.com/Evan-Kim2028/igloo-mcp/discussions)*