Let Claude Code build data artifacts in DataUnmess.

Use the guided setup page to create a key, copy the Claude Code instructions, save the MCP config, replace the placeholder with your key, and restart Claude Code.

Claude Code is useful when the data workflow connects repo context, SQL, Python, and DataUnmess artifacts. It can create pipelines, inspect failures, and explain dashboard source paths.

Claude Code can combine repository context with DataUnmess source metadata and artifact tools. Ask it to inspect the available fields first, explain the intended query or transformation, and then build the dashboard or pipeline in the active workspace.

For dashboards, require source-backed cards with the dataset and query or connection metadata needed for reload. For pipelines, validate on sample rows before a full run and keep the destination explicit. This produces a reviewable artifact instead of a chart or dataframe that exists only in terminal output.

The MCP connection is scoped by the DataUnmess key. Database and SaaS credentials remain in authorized DataUnmess connections rather than being pasted into the Claude Code conversation.

Give Claude Code an acceptance check with the request: the expected row grain, date coverage, KPI definition, refresh behavior, and one total to reconcile. Clear checks make it easier to review both the generated implementation and the saved DataUnmess result.

• Inspect schemas and source fields before selecting metrics.
• Create a pipeline when cleaning or joins should be reused.
• Build dashboard cards only after the source query is defined.
• Use lineage and run logs to debug the saved result later.