Docs

Query an Embedded MCP result set

Search, filter, page, or aggregate a stored Embedded MCP action result without another provider call.

query-result-set reads a result that execute-actions stored after a large response. It does not call the app again. Get resultSetId from meta.compaction.resultSet.

{"resultSetId":"<result-set-id>","q":"Acme","fields":["id","name"],"limit":25}
{"success":true,"data":[{"id":"row_1","name":"Acme"}],"meta":{"matched":1,"nextCursor":null}}

q searches across rows. filters matches full values by dot-path; string comparisons ignore case. fields selects dot-paths, and sortBy orders rows. limit is 1 to 200, with 25 as the default. Pass meta.nextCursor as cursor for the next page.

Use aggregate for counts or metrics without downloading every row:

{"resultSetId":"<result-set-id>","aggregate":{"groupBy":["status"],"metrics":[{"op":"count","as":"records"}]}}

Metrics support count, sum, avg, min, and max. Each query renews the 30-minute result lifetime. A result belongs to one end user. If it expires, run the original action again.

count needs no field. Other metric operations need the field to measure. When aggregate is set, fields is ignored.