Register a new database connection (MySQL or Snowflake). AFTER USING: Always call test_datasource to verify the connection works. MySQL config: {host, port, username, password, database}. Snowflake config: {account, username, password, warehouse, database, schema}.

List all registered data sources. FIRST STEP: Always check what data sources are available before doing any work. If empty, guide user to add data sources with add_datasource.

Test connection to a data source. USE: After adding a new data source, when user reports issues, or before starting reconciliation.

Remove a registered data source. CAUTION: This deletes the configuration permanently.

List all databases in a data source. USE: To discover available databases before listing tables.

List all tables in a database/schema. PREFER search_tables if user mentions a specific table name pattern.

Get column definitions (name, type, nullable, primary key). USE: To understand table structure, find numeric columns for aggregates, or find primary keys for sample checks.

Get FULL metadata catalog (all databases, schemas, tables). USE: When starting with a data source to understand what's available. WARNING: Can be slow for large databases - prefer search_tables for targeted queries.

🔍 CRITICAL - USE FIRST when user mentions a table name. Search for tables by name pattern. NEVER assume table names exist - always search first. Example: User says 'orders table' → search pattern 'order' to find 'orders', 'order_items', etc. Then confirm with user which table they mean.

Preview first N rows of a table. USE: To VERIFY you found the correct table before running expensive checks. Show user what data looks like.

⚠️ REQUIRED BEFORE CHECKS - Verify a table exists before running any reconciliation. ALWAYS call this for BOTH source AND target tables before run_row_count_check, run_aggregate_check, run_schema_check, or run_sample_check. Prevents wasted time and cryptic errors.

⚠️ REQUIRED FOR AGGREGATES - Verify columns exist before running aggregate checks. ALWAYS call this before run_aggregate_check to prevent errors from typos or missing columns.

Get quick statistics: row count, column count, size. USE: For sanity check before reconciliation, to estimate how long checks might take (large tables = longer).

📊 RECOMMENDED FIRST STEP - Side-by-side comparison of source and target schemas BEFORE running data checks. Shows column differences and type mismatches that may cause false positives in data comparison.

✅ FAST FIRST CHECK - Compare row counts. Run this FIRST - it's fast and catches major issues. PREREQUISITE: Call validate_table_exists for both tables first. For large tables, use partition_config to compare date ranges.

Compare column-level aggregates: SUM, AVG, MIN, MAX, COUNT_DISTINCT. PREREQUISITES: 1) validate_table_exists 2) validate_columns_exist. USE: After row counts match, to verify numeric data integrity.

Compare table schemas (column names, data types). USE: To identify structural differences that might cause data issues.

🔬 DETAILED CHECK - Compare actual row values by primary key. USE: After row counts and aggregates pass, for detailed validation. Identifies specific mismatched rows. PREREQUISITE: validate_table_exists, know the primary key.

Create comprehensive reconciliation job with multiple checks. Runs asynchronously - use get_job_status to monitor. PREREQUISITES: validate_table_exists, compare_table_structures.

Get job progress: status, percent complete, current check. Poll every few seconds while running. Report to user: 'Job is 50% complete, running aggregate check...'

Get detailed results after job completes. USE: After get_job_status shows status=completed.

List recent reconciliation jobs. USE: To see job history or find previous results.

Cancel a running job. USE: When user wants to stop long-running reconciliation.