Dry-run pricing for an enrich run - THE money-honesty gate (§4.6): call it BEFORE any paid create_playtable_run(type:"enrich") and surface the range to the user. Resolves the scope row count per field (ES run_condition prefilter - approximate by design, the worker re-checks per cell; done cells are skipped unless force_rerun:true = re-charge consent; cells:not_run | error prices only the cells that have not run or only the error ones) and prices each field's materialized waterfall from the cached gtm-api tool registry.

Returns rows_matched, cost_min (every row succeeds on the cheapest step), cost_max (full waterfall at max prices), cost_expected (registry expected path), priced_at, and a per-field breakdown with per-step price ranges - pass cost_max as max_budget at create (cost_expected is a midpoint, not a cap). No writes, no charges - free to call; free run types (export|import|bulk) have nothing to estimate. Also the answer to "what would it cost to …" - estimate only, no run. A view/filter scope is priced by count (rows_matched = the ES count; rows_to_run per field subtracts the rows already done unless force_rerun, or counts the error rows by id under cells), so any size prices instantly and rows_matched is never a truncated number.

Contract:

• MCP tool estimate_playtable_run, registry package mcp.playtables/playtable_runs, mount playtables.columns.
• Operation action, response envelope action.
• Flags: read only.