ONBF
DocumentationGoogle Drive

Google Drive connector

Markdown

When a user connects their Google Drive, your agent gets four read-only tools to list, search, read and download their files — on their behalf, with no configuration on your side.

#Overview

The Google Drive connector lets your agent work with files in a user's own Drive. The user authorizes ONBF once with Google's read-only Drive scope (drive.readonly); ONBF vaults and refreshes the OAuth token, and exposes four tools on the run's session token. You never register a Google app, request scopes or handle a callback — see Connectors for the zero-setup model.

Read-only, per-user: This connector uses Google's drive.readonly scope: your agent can list, search, read and download the user's files, but never create, edit or delete anything. Access is the individual user's — granted and revocable from their ONBF dashboard.

#Tools

These four tools are exposed when the builder selects Google Drive under Required connectors and the run credential holds connectors:use. User connection status is checked when a tool runs.

ToolWhat it doesKey arguments
google_drive_list_filesList the user's Drive files, most-recently-modified first. Returns id, name, type, size and a web link for each.pageSize (optional, 1–100), pageToken (optional cursor).
google_drive_search_filesSearch the user's Drive by a plain-text query (matches file names and full-text content). Returns matches newest-first with id, name, type and a web link.query; pageSize (optional, 1–100), pageToken (optional cursor).
google_drive_read_fileRead the text contents of one file by its id. Google Docs export to plain text and Sheets to CSV; other text files return as-is. Large files are truncated.fileId — an id from list/search.
google_drive_download_fileDownload one file so the agent can process it locally (PDF, image, spreadsheet, archive, …). Returns a short-lived downloadUrl — NOT the bytes.fileId — an id from list/search. exportMimeType (optional) — override the export format for Google-native files.

#read_file vs download_file

Two tools return file data, and picking the right one matters. google_drive_read_file is for text you want inline — it returns the file's text content directly in the tool result (Docs as plain text, Sheets as CSV). google_drive_download_file is for binaries you need on disk — it returns a link, not the content.

download_file returns a URL, not bytes: File bytes never ride through an ONBF Tool call. google_drive_download_file returns a short-lived, purpose-bound downloadUrl; fetch it with a plain HTTPS GET and no auth header. The user's Google token stays server-side and is never exposed to the agent.

Google-native files (Docs, Sheets, Slides) have no raw binary, so google_drive_download_file exports them automatically:

Google-native typeDefault export
Google DocsPDF
Google Sheets.xlsx (Excel)
Google SlidesPDF

Pass exportMimeType to override the default (e.g. export a Doc as text/plain or a Sheet as text/csv). For ordinary binaries (PDF, PNG, ZIP, …) exportMimeType is ignored — you always get the original bytes.

There's a size limit: Downloads are capped at 25 MB. Larger files return a clear error instead of a link — narrow the selection or prefer google_drive_read_file for text-heavy documents.

#Typical flow

  • Discover — list_connections confirms the user has google-drive connected and returns the exact tool names.
  • Find the file — google_drive_search_files (by query) or google_drive_list_files (browse), keeping the returned id.
  • Read or download — for text, google_drive_read_file returns the content inline. For a binary, google_drive_download_file returns a downloadUrl to fetch over HTTPS.
  • Process it — parse, summarize or transform the file, then deliver the result back to the user as a reply or an artifact.
Google Drive connector · ONBF