Gmail MCP connector: search_threads resultCountEstimate wildly wrong for in:inbox

Summary

mcp__Gmail__search_threads with query in:inbox returns a resultCountEstimate far higher than the actual number of threads in the inbox.

Steps to reproduce

1. Call search_threads with query: "in:inbox", any pageSize.
2. Compare resultCountEstimate in the response to the actual inbox count shown in the Gmail web UI (the "1-XX of YY" text above the message list).

Observed

• search_threads returned resultCountEstimate: "201".
• The Gmail web UI, checked via an authenticated browser session against the same account, showed 1-24 of 24.

That's roughly 8x off, not a rounding/estimate artifact.

Expected

resultCountEstimate for in:inbox should be reasonably close to the actual inbox thread count, or the tool description should warn explicitly that this field is unreliable for in:inbox queries specifically (it currently just says results "exceeding pageSize" trigger an estimate, which reads as roughly-accurate).

Impact

An agent using this connector to answer "how many emails are in my inbox" or similar will confidently report a number ~8x too high unless it happens to cross-check against another surface (in this case, a screenshot of the actual Gmail UI caught it).

Environment

• Gmail MCP connector (mcp__Gmail__*), accessed via Cowork
• Verified same-session, same account, within minutes of each other