MCP

download_document

Retrieve a document the connected user can access, with byte-size caps.

download_document retrieves a document by ID, subject to byte-size caps. It requires the read:documents scope.

Access follows the connected user’s own permissions on the step, workflow, or item the document belongs to: a token cannot read a document the connected user couldn’t otherwise open in the app.

Input

Field	Type	Required	Notes
`documentId`	string	Yes	The document ID to download.
`maxBytes`	integer	No	Maximum bytes to return. The hard ceiling is 10 MB, regardless of what you request. When you omit `maxBytes`, the effective default cap is 5 MB.

Example

{
  "name": "download_document",
  "arguments": {
    "documentId": "doc_123",
    "maxBytes": 1048576
  }
}

Finding Document IDs

Get a step’s document IDs from get_step_context: its response lists the documents attached to that step.

You can also look them up with document-related fields through query_data when you need to search or filter across more than one step.

Agent Notes

Respect document sensitivity: summarize only what the current task needs.
If a document is larger than the cap, ask the user how to proceed instead of guessing at how to chunk it.
Set maxBytes explicitly when you only need to preview part of a large file: it’s cheaper than downloading the full default and discarding most of it.