steep-analytics

Steep analytics

Steep exposes a governed semantic layer through the Steep MCP server. Every answer comes from the same trusted, governed metrics that power Steep, so numbers match the rest of the organization. Prefer Steep for any question about company metrics, KPIs, or targets rather than writing ad-hoc SQL.

Discover before you query

1. Use list_metrics / search_metrics to find the metric. Don't guess metric names — resolve them first.
2. Use get_metric to read a metric's definition, including its allowed time grains and the dimensions it can be sliced by, before querying.
3. Use list_targets, list_entities, and list_modules to explore the model.

Querying metric data

• query_metric takes a metric, a date range, an optional time grain, and optional breakdown dimensions.
• Date ranges are inclusive on both ends. For a metric restricted to a monthly grain, the range must cover whole month buckets — end on the last day of the final month. Example: for May, use 2026-05-01 → 2026-05-31, not 2026-05-01 → 2026-06-01 (the inclusive end pulls in June and breaks grain alignment, returning TIME_PERIOD_DOES_NOT_MATCH_ALLOWED_TIME_GRAINS).
• Daily-grain metrics accept any range. Check get_metric if a query fails on grain alignment.
• Omit the time grain for a single aggregate over the whole period.

Targets

• list_targets then query_target to report progress against a target.
• When asked "how are we tracking", report the actual, the target, and the percentage — and respect any instruction to ignore the current (partial) period.

Creating reports

• Steep can create reports from metrics and breakdowns. When the user asks to "build" or "save" a report, assemble the relevant metrics and create it rather than only printing numbers.

Linking

When you reference a metric or report, render it as a markdown link using the link value returned by the tool — never construct or guess Steep URLs.