Igloo MCP

# search_catalog Query the catalog artifacts produced by `build_catalog` and retrieve matching objects (tables, views, functions, etc.) along with basic metadata and columns. ## Description Use this tool after running `build_catalog` to explore the locally cached snapshot without reconnecting to Snowflake. It supports substring filtering on object names or column names and returns at most `limit` matches. **Unified Storage**: When `catalog_dir` is the default, the tool resolves to unified storage location (`~/.igloo_mcp/catalogs/{database}/`). Set `search_all_databases=true` to search across all database catalogs in unified storage. ## Parameters | Name | Type | Required | Default | Description | |------|------|----------|---------|-------------| | `catalog_dir` | string | ❌ No | `./data_catalogue` | Directory containing `catalog.json` or `catalog.jsonl`. Default resolves to unified storage at `~/.igloo_mcp/catalogs/{database}/` when used with `build_catalog` default. | | `search_all_databases` | boolean | ❌ No | `false` | If `true` and `catalog_dir` is default, search across all database catalogs in unified storage. | | `object_types` | array[string] | ❌ No | — | Optional list of object types (`table`, `view`, `function`, …). | | `database` | string | ❌ No | — | Filter results to a specific database. | | `schema` | string | ❌ No | — | Filter results to a specific schema. | | `name_contains` | string | ❌ No | — | Case-insensitive substring match on object name. | | `column_contains` | string | ❌ No | — | Case-insensitive substring match on column name. | | `limit` | integer | ❌ No | 20 | Maximum number of results to return (1-500). | | `request_id` | string | ❌ No | auto-generated | Request correlation ID for distributed tracing (UUID4). Auto-generated if not provided. | ## Example ```json { "status": "success", "request_id": "650e8400-e29b-41d4-a716-446655440001", "catalog_dir": "./artifacts/catalog", "total_matches": 3, "limit": 20, "results": [ { "object_type": "table", "database": "ANALYTICS", "schema": "REPORTING", "name": "SALES_FACT", "columns": [ {"name": "ID", "data_type": "NUMBER"}, {"name": "REGION", "data_type": "VARCHAR"} ] } ], "timing": { "search_duration_ms": 12.34, "total_duration_ms": 15.67 }, "warnings": [] } ``` ### Response Fields - **`request_id`**: UUID4 correlation ID for distributed tracing - **`timing`**: Performance metrics in milliseconds - `search_duration_ms`: Time spent searching catalog data - `total_duration_ms`: Total execution time - **`warnings`**: Array of non-fatal issues (e.g., missing catalog files, empty if none) ## Notes - If the catalog directory does not contain a build artifact, the tool returns an error instructing the user to run `build_catalog`. - Column metadata is surfaced when available so downstream tools can inspect schemas without rerunning catalog queries.