How long a document can it handle?

Reports of around 100 pages go to Claude in one request. Longer documents are split into chunks; each chunk is briefed on its own, then Claude merges the chunk briefs into one. Chunk size is a setting, and the guide explains the trade-off between cost and detail.

What about scanned PDFs?

A scanned PDF has no text layer, so the extract step returns nothing and the workflow posts it to Slack as unread instead of guessing. The guide includes an optional branch that sends the PDF itself to Claude through the Anthropic API's document input, which reads each page as an image. It works on scans and costs more per page.

Can it write the brief to Notion or SharePoint instead?

Yes. The Google Docs node is one step. Swapping it for Notion's create-page or the SharePoint file-upload operation is a one-node change, and the guide shows both. The Drive trigger and the Claude step are unchanged.

Does it need a public URL or a paid n8n plan?

No public URL. The Drive trigger polls, so a self-hosted n8n behind a firewall works. Any n8n Cloud plan is enough; a folder that receives 20 PDFs a month uses 20 executions.

Will it summarise files already in the folder when I turn it on?

No. The trigger only sees files created after activation. The package includes a second, manually run workflow that takes a folder and briefs every PDF in it once.

DrCGDSMade to order

AI drafting, summaries & agents

Google Drive PDF → Claude one-page brief → Google Doc + Slack post

Drop a report in the folder; a one-page brief and a Slack link exist minutes later.

Long PDFs pile up faster than anyone reads them: supplier reports, tender documents, research papers, policy drafts. Drop one in a watched Google Drive folder and Claude writes a one-page brief of the key points, the figures that matter and the open questions. The brief is saved as a Google Doc beside the original and its link posted to Slack. Analysts, consultants and paralegals read the brief first and open the source only when they need to.

How it works

Dr
Trigger
New file created in a watched folder (polled every minute)
Google Drive
if
Step 1
Accept PDFs only
Checks the MIME type and size. PDFs up to the limit you set continue; anything else is left in the folder untouched, with a line in the execution log.
Dr
Step 2
Download the file
The Google Drive node downloads the PDF as binary data for the next step. Nothing is moved or renamed in the folder at this point.
if
Step 3
Extract the text
n8n's Extract from File node pulls the text layer out of the PDF and reports the page count. A scanned PDF with no text layer is flagged and skipped; the FAQ covers the option for scans.
C
Step 4
Write the brief with Claude
The full text, the filename and your brief template go to Claude in one request. Back comes a summary in fixed sections: purpose, key points, figures and dates, risks or open questions, and a suggested next step.
GD
Step 5
Create the brief as a Google Doc
A new Google Doc named Brief - <original filename> is created in the same folder. Each section sits under its own title line, and a link back to the source PDF goes at the top.
S
Step 6
Post the link in Slack
One message to the channel you choose: the PDF name, who uploaded it, the first two key points, and links to both the brief and the original.

What changes once it runs

Each brief lands in the same Drive folder under the original's name, on the first poll after the upload finishes.
Fixed sections in every brief, so a reader knows where the figures and the open questions sit before opening it.
Slack carries the link and the first two points; the folder stays the place of record.
Scanned PDFs with no text layer are named in Slack as unread rather than summarised badly.
The brief template is a text file in the package. Section names and length are yours to change.

What you get

Tested n8n workflow export, run end to end on real PDFs
The Claude prompt and brief template as separate files, with the section list marked for editing
A setup guide for the Google OAuth client, the Anthropic key and the Slack app
A recording of one PDF going from folder to brief to Slack
A note on document size limits and what to do with scanned PDFs
Setup support by email for thirty days after purchase

Before you start

Google Workspace or personal Google account with Drive and Docs, OAuth connected in n8n
Anthropic API key with a small prepaid balance
A Slack workspace where you can install an app with the chat:write scope
An n8n instance, Cloud or self-hosted, on a current 1.x release for the Extract from File node
One Drive folder to watch and one Slack channel to post to

Built for

Questions before you buy

How long a document can it handle?: Reports of around 100 pages go to Claude in one request. Longer documents are split into chunks; each chunk is briefed on its own, then Claude merges the chunk briefs into one. Chunk size is a setting, and the guide explains the trade-off between cost and detail.
What about scanned PDFs?: A scanned PDF has no text layer, so the extract step returns nothing and the workflow posts it to Slack as unread instead of guessing. The guide includes an optional branch that sends the PDF itself to Claude through the Anthropic API's document input, which reads each page as an image. It works on scans and costs more per page.
Can it write the brief to Notion or SharePoint instead?: Yes. The Google Docs node is one step. Swapping it for Notion's create-page or the SharePoint file-upload operation is a one-node change, and the guide shows both. The Drive trigger and the Claude step are unchanged.
Does it need a public URL or a paid n8n plan?: No public URL. The Drive trigger polls, so a self-hosted n8n behind a firewall works. Any n8n Cloud plan is enough; a folder that receives 20 PDFs a month uses 20 executions.
Will it summarise files already in the folder when I turn it on?: No. The trigger only sees files created after activation. The package includes a second, manually run workflow that takes a folder and briefs every PDF in it once.

Nearby on the shelf

Whole catalog