Snowflake Warehouse Cost Optimizer

What it does

Snowflake Warehouse Cost Optimizer is a FinOps agent for data engineering and cost-management teams. It queries ACCOUNT_USAGE.WAREHOUSE_METERING_HISTORY and ACCOUNT_USAGE.QUERY_HISTORY to surface idle compute (4+ hours inactive) and the top 20 most expensive queries by execution_time times credits_used, then produces prioritized recommendations and estimated monthly savings (with a plus/minus 15% margin disclaimer). Output can be written to Google Sheets and posted to a Slack channel. Guardrails ensure no warehouse is ever automatically suspended — all actions require manual review, and production warehouses (prod_* prefix) are always downgraded to watch-only status. Built for data platform and FinOps teams who want automated Snowflake cost visibility instead of manually querying ACCOUNT_USAGE views.

Example prompts

• Run the full Snowflake cost analysis and post it to Slack
• Which warehouses are idle and should be auto-suspended?
• Show me the most expensive queries this month
• Estimate monthly savings from idle warehouse credits

Tools (8)

Tools the agent exposes — your AI client calls them automatically when it needs them.

• run_full — Run the Snowflake cost optimizer end-to-end: collect ACCOUNT_USAGE data, analyze idle warehouses and expensive queries, synthesize an LLM narrative.
• analyze_data — Given a RawData object, run pure deterministic analysis and return scored findings, recommendations, and savings estimates. No LLM.
• list_warehouses — List Snowflake warehouses from provided metering data with idle status, credit usage, and auto-suspend recommendations.
• flag_queries — Identify expensive queries from provided QUERY_HISTORY data, ranked by cost score.
• estimate_savings — Compute estimated monthly savings in USD for a list of warehouses.
• list_capabilities — List the agent's static capabilities, credential slots, and guardrails.
• plan_inputs — Plan/brainstorm the inputs for a tool: returns questions, schema and a ready-to-edit example.
• discover_intent — Understand your goal and co-design the exact input via clarifying questions before running.