Please add run_query / get_query to the Looker remote MCP server

Looker Q&A Looker
,
1

Summary

The Looker MCP server currently exposes query tools that are spec-based only (query, query_url, query_sql all take model + explore + fields + filters). There is no tool that can operate on an existing saved Query identifier — i.e. the qid found in an explore URL (.../explore/<model>/<explore>?qid=<slug>). We’d like tool(s) that can resolve and/or run a query given its id/qid, wrapping the Looker API endpoints that already support this.

Use case

Our agent frequently receives Looker explore URLs from users and from saved content (e.g. https://<instance>/explore/my_explore/bigquery_usage_stats?qid=AW7N0llsrGvpUtRWWUDUnZ). Today, because the qid is opaque and no MCP tool can resolve it, the agent cannot return the data behind that link — it has to ask the user to re-open the URL in Looker and paste an expanded fields=…&f[…] URL, or rebuild an approximate query from scratch. This is a poor experience and can’t reproduce the exact saved query.

2

Good request. Supporting get_query and run_query by query ID would let agents reproduce the exact query behind a shared Explore URL instead of reconstructing it from fields and filters. It would be especially useful if the tool accepted both the numeric query ID and the opaque qid, resolving the latter before execution while preserving the current user’s permissions.