Lite: MCP tool coverage gaps — 8 collectors with no tools

[erikdarlingdata]

Problem

8 DuckDB tables have collected data but no MCP tool to expose it. The most critical gap is query_snapshots — when the analysis engine detects a CPU spike, there's no MCP tool to answer "what query was running at that time?"

Missing Tools

Priority	Table	What it contains	Suggested tool
High	query_snapshots	Active queries from sp_WhoIsActive	get_active_queries (with time window filter)
High	server_config	sp_configure key/value pairs	get_server_config
High	database_config	RCSI, recovery model, auto-shrink, etc.	get_database_config
Medium	database_size_stats	Database/file sizes, volume free space	get_database_sizes
Medium	server_properties	Edition, version, CPU count, memory	get_server_info
Medium	session_stats	Connection counts, running/sleeping	get_session_stats
Low	trace_flags	Active trace flags	get_trace_flags
Low	database_scoped_config	Database-scoped config values	get_database_scoped_config

database_scoped_config is also not read by the analysis engine — complete blind spot.

Context

Found during ErikAI testing: analyze_server detected a CPU spike to 99% and recommended get_top_queries_by_cpu, but that tool shows plan cache aggregates, not point-in-time snapshots. The actual query causing the spike was only findable by querying query_snapshots directly via Python.

Acceptance Criteria

• Each table above has a dedicated MCP tool returning its data
• Tools support server_name and hours_back parameters
• get_active_queries supports time window filtering (start/end or hours_back)