Ask the agent things like:
- *"How many PDFs has r3mcnett@gmail.com created?"* - *"List every document Alice uploaded this quarter."* - *"What spreadsheets does bob@acme.com own?"* - *"What sensitive docs did our former contractor at carol@example.com create before they left?"*
Behind the scenes the agent calls the **listDocumentsByCreator** MCP tool. You can give it either an email or a userId; if you've been chatting and already named the person, the agent re-uses what it has.
**MIME family vs. exact type.** Say "PDFs" / "spreadsheets" / "images" / "Office docs" / "emails" and the agent uses the MIME-family vocabulary — the same one /search and saved searches use. Say "DOCX" or any other exact type and the agent passes the literal mime type instead.
**The count is permission-aware.** The total returned never includes docs you don't have read on. If you're not an admin, the count reflects only the documents you're entitled to see — Kodori never leaks "there are 47 confidential PDFs you can't see" information through a count.
**Common use cases:**
- *Offboarding review.* "What documents did <former teammate's email> create in the last 30 days?" — pair with retention class changes before access is revoked. - *Spot-checking activity.* "How many docs has Alice uploaded this week?" — leading indicator for a busy matter. - *Audit prep.* "List every regulated-tier doc bob@acme.com created" — narrow first, then run the agent to summarize.
**What this is NOT.** This counts documents **created** by someone — that's the upload + email-ingest + Office-add-in path. It does not yet count documents someone last **modified** (i.e. uploaded a new version of) — that's a different attribution and lives on document_versions.authorId. If that distinction matters for your audit, ask for it in /feedback and we'll layer in a last-modified variant.