Skip to main content
This document covers import_from_dropbox (display name Dropbox Import), which copies one Dropbox file into platform storage for use by downstream extraction, coding, and file tools. Dropbox is a native REST integration, not an MCP or Composio tool.

​
When to use

Use this tool for an on-demand, point-in-time import of one known file. Use a Dropbox file-import trigger instead when a task should start whenever a watched folder or any of its subfolders receives a new file.

​
Authentication and enablement

The tool uses the Dropbox OAuth connector. The access token is injected through x-variable-service: "dropbox" and is not shown as a per-run input. Dropbox API calls use the audited HTTP transport. Off by default. Enable the tool explicitly for each agent that needs Dropbox access. Dropbox Business accounts resolve root_namespace_id and send Dropbox-API-Path-Root, so paths can address the team space instead of being silently restricted to the member’s home namespace.

​
Inputs

Provide exactly one source selector:
  • file_path: absolute Dropbox path, such as /Reports/report.xlsx.
  • file_id: stable Dropbox ID beginning with id:.
  • shared_link: HTTPS Dropbox shared URL available to the connected account.
  • revision: hexadecimal Dropbox revision ID for a historical file version.

​
Limits and behavior

  • Imports are limited to 250 MB, checked from metadata and while streaming.
  • The file is streamed through a temporary file rather than placed wholly in later steps.
  • Empty files are rejected explicitly.
  • Shared links must use a Dropbox-owned HTTPS host.
  • The imported copy does not track later Dropbox changes; import it again to refresh the platform copy.

​
Watched-folder imports

Dropbox file-import triggers watch the selected folder recursively and deduplicate each normal file by stable Dropbox ID plus revision. For legacy workflows, enable Legacy combined automated-folder import. This preserves the feature-flagged legacy layout: selected root / level 1 / level 2 / automated / supported files The poller runs no more frequently than every 15 minutes, waits until the automated folder has been quiet for 15 minutes, converts PDF, PNG, JPG, JPEG, DOC, and DOCX inputs to PDF, and creates one combined-PDF task named after the level-2 folder. The stable level-2 Dropbox ID is the deduplication identity, so the folder is imported once, matching legacy behavior. DOC/DOCX conversion requires the platform’s existing ConvertAPI configuration.

​
Output

The structured result contains file_id, file_name, dropbox_id, dropbox_path, shared_link, revision, storage_url, mime_type, file_size, and completed_at. A workspace-scoped files row is created when task execution context is available. In React tasks, storage_url may be returned as a task-variable reference so the raw storage URL stays out of model context.

​
Errors

  • Missing or multiple selectors fail validation before a Dropbox call.
  • A path, ID, or shared link that identifies a folder is rejected.
  • Malformed IDs, revisions, and non-Dropbox shared links are rejected.
  • Empty or oversized content is not stored.
  • Dropbox and storage failures return structured tool errors without logging tokens, paths, links, response bodies, or file content. The error carries the HTTP status plus, when Dropbox reports one, the union tag (path/not_found) or the name of a scope the app has not been granted.
  • A Dropbox app missing a scope fails every file call with a 400. The error names the scope; grant it on the Permissions tab of the Dropbox App Console and reconnect the integration, because Dropbox narrows the grant at authorization time and existing tokens keep the old, narrower scope set.