Back to catalog / Agent workflows

Query supply-chain data and produce stockout actions

Use Databricks MCP to query structured tables, emails, and model results; answer inventory/supply questions with data evidence.

Goal

Produce an evidence-backed stockout-risk and human-action list for specified SKUs and dates.

Run notebook 1 of the Databricks Solution Accelerator in a test workspace and configure ~/.databrickscfg plus an OpenAI key. [src10-openai-query-supply-chain-data-for-stockout-actions]
Ask an inventory/supply question; the agent queries tables, emails, and models through MCP and retains source tables/records for each claim. [src10-openai-query-supply-chain-data-for-stockout-actions]
Summarize verified risks, SKUs/dates, and suggested human follow-up; do not invoke write tools to place orders. [src10-openai-query-supply-chain-data-for-stockout-actions]

Verify SKUs against Databricks tables, emails have thread/date evidence, model outputs show version, and no action claims an order was placed.

For MCP failure, check Databricks profile and warehouse permissions; for unsupported claims, specify SKU/period instead of inventing stock.

Databricks test workspace and accelerator assets, ~/.databrickscfg, and an OpenAI API key.

Reads Databricks tables, emails, and model outputs and sends questions to OpenAI; the sample does not create purchase orders.

The accelerator creates test data assets and may incur charges; output flags risk, not supplier default or purchase commitment.

Repository: openai/openai-cookbook
File path: examples/mcp/building-a-supply-chain-copilot-with-agent-sdk-and-databricks-mcp/README.md
Pinned commit: 5bd6cb40b6b84b69b2032d810610aa85064102df
Source checked: 2026-10-14 · Reviewed against the pinned upstream source

These values describe the upstream repository, not popularity or likes for an individual skill or playbook.

No metrics snapshot is available yet.

License: MIT