SQL warehouse as agent context
Let agents query production analytics (SQL, Iceberg/DuckLake, shared semantic Guides) rather than dumping a warehouse into a vector DB.
At a glance — Oct 2026
Updated 2026-10-27 · forward this page: /jtbds/sql-warehouse-agent-context#briefScores as of 2026-10-19 (8 days old).
- Depot Pick
- MotherDuck · 8.44/10 for this job (curated estimate, high confidence)
DX for programmatic SQL + pricePerf of bursty agent queries. MotherDuck Lite/Pulse is the agent-native default; hyperscaler warehouses when the data already lives there.
- Runner-up
- Snowflake Cortex · 7.10/10
Scores higher than the pick on: Maturity / track record.
- When not to use the pick
- Snowflake is already the system of record — Cortex; Databricks lakehouse standard — Mosaic; local/embedded only — DuckDB OSS.
- Lock-in / migration
- The pick has lower stored lock-in risk than runner-up Snowflake Cortex (4.5 vs 8). Self-host fallback listed for this job: DuckDB.
- Sourced price (no comparable unit)
Pick: no sourced price in this job's unit (no_matching_sourced_price).
No cheapest SKU (no_comparable_unit_for_job). No price was invented.
- Free option
- DuckDB (oss) — source, as of 2026-10-25
- Failure modes to watch
- Schema drift: Validate every tool call and every loaded table against the current schema. Quarantine mismatches instead of coercing them.
- Collateral side effects: Snapshot the workspace or app state, then diff it against an allowlist. A passing check plus an unexpected diff is a failure.
- Hallucinated tool arguments: Reject a call that fails the current schema or names an unknown id, and do that before the tool runs a side effect.
- Partial tool success: Require a per-item status and a failure count. Success for the task requires the failed count to be zero or explicitly handed to a retry queue.
- Silent truncation: Tools must return whether the payload was truncated and how long the original was. Fail the task if the answer depends on the missing span.
- Stale retrieval: Store the source URL and the time it was fetched. If a newer canonical page exists, or two stored sources disagree, do not answer from the older one.
Cite: Agent Provider Depot, Depot Pick for sql-warehouse-agent-context: motherduck, scores as of 2026-10-19. Agents: GET /api/decide?jtbd=sql-warehouse-agent-context
Free SQL warehouse as agent context options · for data-etl agents
MotherDuck
DX for programmatic SQL + pricePerf of bursty agent queries. MotherDuck Lite/Pulse is the agent-native default; hyperscaler warehouses when the data already lives there.
Pricing (directional): As of 2026-10-24 (motherduck.com/product/pricing, fetched 2026-10-24): Lite $0 — 3 users, 2 service accounts, 10 GB, 10 Pulse hrs/mo. Business $250/org/mo + usage — 10 users, unlimited service accounts, 99.9% SLA. Enterprise custom (Private…
Lock-in risk: higher = more lock-in (worse). Overall already inverts this dimension.
Depot pick when the agent needs SQL warehouse context (analytics, DuckLake/Iceberg, shared Guides). Snowflake Cortex / Databricks Mosaic when the lakehouse is already the system of record.
Quote for SQL warehouse as agent context
Depot Pick stays MotherDuck. Prices below are sourced SKU rows only (as of 2026-10-27). Input and output are not blended. GET /api/quote?jtbd=sql-warehouse-agent-context
Nothing was compared. no comparable sourced unit for this job; chat input $/1M and other units were not borrowed pickIsPriceWinner is null.
- AI Functions usd_per_ai_unit $1 source as of 2026-10-03
- Business usd_per_org_month $250 source as of 2026-10-03
- Flight F16 usd_per_hour $0.6 source as of 2026-10-03
- Flight F32 usd_per_hour $1.2 source as of 2026-10-03
- Flight F4 usd_per_hour $0.15 source as of 2026-10-03
- Giga usd_per_hour $24 source as of 2026-10-03
- Jumbo usd_per_hour $4.8 source as of 2026-10-03
- Lite usd_per_month $0 source as of 2026-10-03
- Mega usd_per_hour $12 source as of 2026-10-03
- Pulse usd_per_cu_hour $0.6 source as of 2026-10-03
- Standard usd_per_hour $2.4 source as of 2026-10-03
- Storage usd_per_gb_month $0.04 source as of 2026-10-03
4 of 5 candidates have no sourced price row. 5 are excluded from the Matching price comparison (none). no comparable sourced unit for this job; chat input $/1M and other units were not borrowed
Runners-up
Snowflake Cortex
Use when Snowflake already holds the data. Don't drag agent backends into Snowflake just for Cortex.
Databricks Mosaic AI
Pick only when your data/platform already is Databricks. Otherwise Bedrock/Vertex or direct APIs.
DuckDB (vector/extensions)
Use for local analytics + light vector experiments. Production agent memory → pgvector/Qdrant/Mem0.
ClickHouse Cloud
Depot pick for agent-analytics-warehouse (event/log OLAP). MotherDuck remains sql-warehouse-agent-context for…
When NOT to use the top pick
Snowflake is already the system of record — Cortex; Databricks lakehouse standard — Mosaic; local/embedded only — DuckDB OSS.
Full ranking (score table)
Ranked with this JTBD's weights. Scores are curated estimates unless marked otherwise. Agents: prefer /api/decide then this table for tie-breaks.
| # | Provider | Overall | Fit | $/perf | Latency | DX | As of |
|---|---|---|---|---|---|---|---|
| 1 | MotherDuckDepot Pick | 8.4 | 8.8 | 8.6 | 8.6 | 9.0 | 2026-10-19 |
| 2 | ClickHouse Cloud | 8.4 | 9.0 | 8.0 | 8.8 | 8.4 | 2026-10-19 |
| 3 | DuckDB (vector/extensions) | 8.0 | 6.0 | 9.5 | 7.5 | 8.5 | 2026-09-14 |
| 4 | Databricks Mosaic AI | 7.1 | 8.4 | 4.8 | 8.5 | 6.8 | 2026-09-14 |
| 5 | Snowflake Cortex | 7.1 | 7.8 | 5.5 | 8.4 | 7.0 | 2026-09-14 |
Score methodology for this job
DX for programmatic SQL + pricePerf of bursty agent queries. MotherDuck Lite/Pulse is the agent-native default; hyperscaler warehouses when the data already lives there.
capabilityFit: 25%pricePerformance: 18%latencyReliability: 15%developerDx: 15%privacyPosture: 10%lockInRisk: 7%maturity: 10%