Skip to main content
This document covers download_google_drive_link (display name “Google Drive Download from Link”). Given a Google Drive link, it downloads that file — or every file directly inside that folder — into the running task, and returns a file_url per file that downstream tools (structured_data_extraction, csv, python_code, export tools) accept.

​
When to use — and when not

Use it when an agent is handed a Drive link and needs what is behind it now — e.g. “process the invoices in this folder: https://drive.google.com/drive/folders/…”.
  • To find one file by name without a link, use search_and_download_google_drive_file.
  • To process files as they arrive in a folder, use a Google Drive file_import trigger. The trigger only picks up files added after it is set up; this tool is how you import what is already there.
The download is a point-in-time copy; later changes in Drive are not reflected. Call again to refresh.

​
Authentication

Uses the workspace’s Google Drive connection (the first active one). The access token is injected by the platform and is not a per-run input. The folder or file must be visible to the connected Google account — shared with it, in its My Drive, or in a shared drive it belongs to. Outbound Drive API calls are audited. Off by default — enable it in the agent’s Tools panel.

​
Inputs

  • link (required): a folder link (drive.google.com/drive/folders/<id>, any /u/<n>/ variant), a file link (/file/d/<id>, open?id=, uc?id=, docs.google.com/.../d/<id>), or a bare Drive ID.
  • include_types (optional): file extensions to keep, e.g. ["pdf", "xlsx"]. Case-insensitive; a leading dot is allowed. Omit to keep every type.
  • max_files (optional): how many files to download from a folder. Default and maximum 10.

​
Behaviour

  • A folder link downloads the files directly inside it, in name order. Subfolders are not opened.
  • A file link downloads that one file.
  • Native Google files (Docs, Sheets, Slides, Forms, shortcuts, …) are skipped — they have no file content to download and are not exported.
  • Files whose extension is not in include_types are skipped.
  • Each file can be at most 25 MB, and at most 100 MB is downloaded per call. A file larger than 25 MB is skipped (or, if Drive under-reported its size, stopped mid-download) as too_large; files that would push the call past 100 MB are skipped as over_size_budget.
  • Links with a ?resourcekey= (some older “anyone with the link” items) are supported: the key is sent back to Drive, along with each file’s own key inside such a folder.
  • Only the first 1,000 items of a folder are considered (folder_truncated: true when there are more).
  • Files are downloaded four at a time. One failed file does not stop the others.
  • Downloading the same file (same name and size) twice in one task reuses the copy already saved.

​
Output

Structured result:
  • source: id, name, type (folder | file), url of the linked item.
  • files: every file a download was attempted for — name, mime_type, size, drive_url, saved, and when saved file_url and file_id (the platform files row); otherwise not_saved_reason (too_large, empty_file, storage_unavailable, download_failed).
  • saved_count, failed_count.
  • skipped: up to 50 items not attempted, each with name and reason (subfolder, google_native_file, type_not_included, over_max_files, over_size_budget, too_large); skipped_count is the full number.
  • folder_truncated when the folder had more than 1,000 items.
On react tasks each file_url is a ${VAR_N} task-variable reference that resolves automatically when passed to the next tool; CodeAct tasks receive the raw storage URL. Each saved file is registered as a task file, so it appears in the task’s files like a manual upload.

​
Errors

The call is reported as failed when:
  • No Google Drive connection is available.
  • link is not a Google Drive link or ID, or no ID can be found in it.
  • The item does not exist or is not shared with the connected account.
  • The folder cannot be listed.
  • A file link points at a native Google file, or at a file outside include_types.
  • Downloads were attempted and none succeeded.
An empty folder, or one where every file was filtered out, is not an error: the result simply has saved_count: 0 and the skipped list explains why.