Runs read-only SQL queries against chosen database connection. Only SELECT queries are allowed. INSERT, UPDATE, DELETE, DROP, and other write operations are blocked. START HERE - REQUIRED DISCOVERY FLOW: - Before writing SQL, read db://{connection} to list available tables for the target connection. - Then read db://{connection}/{table} for the exact schema of tables you plan to query. - This avoids "table not found" and wrong-column errors. - Before querying information_schema/system catalogs for metadata/definitions, try database_schema with detail="full" (and includeRoutines/includeViews when needed). ROW LIMIT: - Default to LIMIT 10 when browsing or exploring data. Use TOP 10 on SQL Server. - Use a larger LIMIT or TOP only when the task requires a specific known row count, such as 31 daily buckets. - Keep the limit as small as the task allows. RULES: 1. SELECT without WHERE requires LIMIT or TOP, except for aggregate-only queries without GROUP BY, such as SELECT COUNT(*) FROM users. 2. Check the connection type before writing the query - use correct syntax for that database. 3. To browse beyond the first page, use pagination with OFFSET. 4. Large text columns (>200 chars) are truncated to "<TEXT>" in multi-row results. To view full text, Query MUST return exactly 1 row. Examples: MySQL/PostgreSQL/SQLite: SELECT * FROM users LIMIT 10; SQL Server: SELECT TOP 10 * FROM users; Aggregates (no LIMIT needed): SELECT COUNT(*) FROM users;

Inspect schema for a database connection. Use database_schema(detail="full", includeRoutines=true) as the default way to fetch trigger/function/procedure/view definitions. Prefer this over raw information_schema/system catalog queries. Detail levels: - summary (default): matching table names. - columns: matching tables with column types. - full: full table structures (columns, indexes, foreign keys, triggers, check constraints). Object coverage: - tables: columns/indexes/foreign keys/check constraints/triggers (with trigger definitions in full detail). - views: SQL definitions when includeViews=true. - routines: function/procedure definitions when includeRoutines=true and detail="full". Include flags: - includeViews=true adds view names for summary/columns. - includeRoutines=true adds stored_procedures, functions, sequences, and trigger names for summary/columns. - Triggers are included under each table for detail="full" and are not duplicated under routines. Definition preference: - If output already includes a definition field, do not re-query raw catalogs/system tables. Usage guidance: - detail="full" and detail="columns" should be used with a narrow filter whenever possible. - Large full/columns outputs are rejected with ToolUsageError; refine filter or use summary. Filter note: - filter is optional and matches object names (tables, views, procedures, functions, sequences, triggers). - In detail="full", matching routine/view/trigger objects include their definitions in output. - Omit filter (or use filter="") to include all object names. Examples: - Get trigger function body by name: connection="users", filter="trg_users_insert_fn", detail="full", includeRoutines=true - Get view SQL by view name: connection="users", filter="my_view", detail="full", includeViews=true - Get all routines in schema prefix: connection="users", filter="public.", matchMode="prefix", detail="full", includeRoutines=true Routine note: - In PostgreSQL, many routines are exposed as functions (not procedures). If detail="columns" or detail="full" output is too large, the tool returns ToolUsageError and asks for a narrower filter.