docs: document the searchExplorations MCP tool

[igorlukanin]

Summary

• Documents the searchExplorations MCP tool, which searches saved queries — explorations and workbook tabs alike — by name and description and returns each one's sqlQuery, ready to pass to runQuery.
• Adds it to the "Query and discovery" table and updates the tool count.

Test plan

• Table renders with the new row; surrounding guidance reads correctly
• CI must pass

[mintlify]

Preview deployment for your docs. Learn more about Mintlify Previews.

Project	Status	Preview	Updated (UTC)
cubed3	🟢 Ready	View Preview	Aug 13, 2026, 11:39 PM

💡 Tip: Enable Workflows to automatically generate PRs for you.

[claude]

Claude finished @igorlukanin's task in 1m 18s —— View job

---

Review complete — docs-only change, looks good to merge. 0 high, 0 medium, 3 low (2 inline nits + 1 verification ask)

Full review

Todos

• Gather context (read diff)
• Verify tool count and table contents in the doc
• Check terminology against docs-mintlify/CLAUDE.md
• Check for existing Claude review threads (none — nothing to resolve or dedupe against)
• Post review

What's correct

• Tool count is accurate. I counted the tool rows across all five tables in mcp-server.mdx — listDeployments, chat, loadQueryResults (3), Query and discovery (3), Dashboard authoring (5), Data model editing (8), pre-aggregations (2) = 21, matching the updated line 232.
• Terminology matches the style guide. "explorations" and "workbook tabs" line up with the product taxonomy in docs-mintlify/CLAUDE.md (Workbook → Tab, Explore → Explorations). No "Cube Cloud" or other legacy naming introduced.
• Access annotation is consistent. Read-only matches the other query/discovery rows, and the destructive-tools paragraph at lines 240–244 still says "four destructive ones" — correctly unaffected by this addition.
• Table syntax is valid (3 columns, matching the header), and the change lives in the active /docs-mintlify site, not the deprecated /docs.

Findings (all low severity, posted inline)

#	Location	Note
1	docs-mintlify/docs/integrations/mcp-server.mdx:263	No link reference for "explorations" — other rows link [ref-sql-api] / [ref-workbooks] / [ref-dashboards]; /docs/explore-analyze/explore#saving-explorations exists and would fit. Also: the row is listed after runQuery while the prose says to call it first.
2	docs-mintlify/docs/integrations/mcp-server.mdx:268-271	The new guidance paragraph is four lines and slightly editorial, next to the one-sentence searchDataModel/runQuery guidance above it. docs-mintlify/CLAUDE.md asks for the shortest version that answers "how do I use this?" — suggested a tighter rewrite.
3	—	Two factual claims can't be verified from this repo (the MCP server implementation isn't here): that the returned field is literally sqlQuery, and that results are access-scoped per user. Both are load-bearing for a reader; worth a confirmation from the implementation.

Not applicable

Security, performance, and test-coverage checks don't apply — this is a 7-line documentation change with no code, no build, and no test surface. I did not run the Mintlify dev server (no dependency install in this environment); the table and MDX are syntactically plain and render-safe by inspection.

• [`igor/cub-3760-document-searchexplorations-mcp-tool`](https://github.com/cube-js/cube/tree/igor/cub-3760-document-searchexplorations-mcp-tool)

[claude]

Claude finished @igorlukanin's task in 1m 11s —— View job

---

Re-review of the updated diff — both prior nits addressed, docs-only, good to merge. 0 high, 0 medium, 1 low (non-blocking)

Full review

Todos

• Gather context (read diff since last review)
• Verify tool count, link reference, and anchor target
• Check existing Claude review threads (both already resolved — nothing to resolve or dedupe)
• Post review

What changed since my last review

Both of my earlier inline nits were addressed, and their threads are already resolved:

1. Link reference added. [ref-explorations]: /docs/explore-analyze/explore#saving-explorations is defined at line 442 alongside the other refs and used in the new table row at line 261. I verified the target: docs-mintlify/docs/explore-analyze/explore.mdx exists, is registered in docs.json, and contains ## Saving explorations (line 50) — so the #saving-explorations anchor resolves.
2. Ordering fixed. searchExplorations now sits at the top of the Query and discovery table (line 261), ahead of searchDataModel and runQuery, matching the "call it first" guidance in the prose below.
3. Concision improved. The editorializing ("the numbers your team already trusts") is gone. The paragraph is still four lines, but the fourth line is now the substantive sqlQuery: null caveat rather than filler, and the Otherwise call searchDataModel… transition reads cleanly.

Verification

• Tool count is correct. Counted the tool rows across all five tables: Deployments and chat (3) + Query and discovery (3) + Dashboard authoring (5) + Data model editing (8) + pre-aggregations (2) = 21, matching line 232.
• Destructive-tools paragraph unaffected. Lines 240–244 still name four destructive tools; searchExplorations is Read-only, consistent with the other rows in its table.
• Terminology matches the style guide. "explorations" and "workbook tabs" line up with the taxonomy in docs-mintlify/CLAUDE.md (Explore → Explorations, Workbook → Tab). No legacy "Cube Cloud" naming introduced.
• MDX is valid — 3 columns matching the header, link reference defined exactly once, change is in the active /docs-mintlify site rather than the deprecated /docs.

Low (non-blocking, no inline comment posted)

Minor tension between the table row and the prose: line 261 says the tool "Returns each one's sqlQuery, ready to pass to runQuery" with no caveat, while line 268 says a saved query with no runnable SQL comes back with sqlQuery set to null. The prose resolves it seven lines later, so a reader isn't left wrong — but if you want the row to stand alone, "Returns each one's sqlQuery, where available" would close the gap. Purely optional.

Still unverified from this repo

The MCP server implementation isn't in this repo, so I still can't confirm two load-bearing claims: that the returned field is literally named sqlQuery (including the null behavior now documented), and that results are scoped to the explorations the caller can access. Both read as intentional and specific, so I'd assume they came from the implementation — flagging only so the claim isn't unverified by silence.

Not applicable

Security, performance, and test-coverage checks don't apply — this is a documentation-only change with no code, no build surface, and no tests. I did not run the Mintlify dev server (no dependency install in this environment); the Mintlify preview build linked above is green.

• [`igor/cub-3760-document-searchexplorations-mcp-tool`](https://github.com/cube-js/cube/tree/igor/cub-3760-document-searchexplorations-mcp-tool)