Summarize Documents with n8n Gateway

Build an n8n workflow that extracts a document or web page and summarizes it.

Introduction

30 Second Summary

A quick AI experiment can lose momentum before the first result appears. Provider sign-ups take time. API keys add another setup hurdle.

In this project, you will build an n8n Cloud workflow that turns a document or webpage into a five-line summary. Gateway credits cover the supported services without provider credentials.

What You'll Build

Picture a PDF becoming a five-line summary in the final node after one workflow run.

By the end of this project, you'll have:

A credential-free AI workflow that produces a model reply without a provider API key.
A document-to-text path that extracts readable content from a PDF or webpage.
An exportable JSON workflow that you can move without exposing secrets.
Secret Mission: An optional challenge to push your skills further.

Are there any prerequisites?

You need an n8n Cloud account that already shows a Gateway credits balance. A browser is the only other requirement. Self-hosted n8n uses a different path.

Before We Start

Your final balance only means something when you have a starting point. This before reading lets you measure what later workflow runs consume.

Your n8n Cloud instance keeps one shared Gateway credits balance. This step gets you signed in with that number recorded.

You also confirm that the workflow editor is ready. That gives you a clear launch point for the first model test.

In this step, get ready to:

Sign in to your n8n Cloud account.
Capture your current Gateway credits balance as the before reading.
Confirm that the workflow editor is available.

Sign in to n8n Cloud

Your Cloud workspace holds the editor used throughout this project. Signing in confirms that your account can reach the tools you need.

Open the n8n Cloud sign-in page in your browser.
Enter the email address associated with your account.
Enter your account password.
Submit the sign-in form.

You should land in your workspace. The left sidebar confirms that the editor interface has loaded.

Having Trouble Signing In?

Use Forgot Password on the login page if you cannot remember your password.
Clear your browser cookies for n8n if the page keeps returning to the login form.
Try another browser if the workspace does not load after authentication.

Need another pair of eyes? Help me troubleshoot my n8n Cloud login. You can also search the n8n Community for account-specific guidance.

Record your starting balance

Checking this balance can feel like a billing step. Opening the balance page does not consume credits.

You are only recording the prepaid amount already shown in the account. This value becomes the baseline for your final comparison.

Select Admin Dashboard in the left sidebar of your workspace.
Select the Gateway credits tab in the Cloud admin dashboard.
Locate the current balance shown on the tab.
Record the amount here: your current Gateway credits balance.

What Does This Number Tell You?

The balance belongs to the whole Cloud instance. Other Gateway credits usage can change it between your two readings.

The project compares this value with the final balance. The difference reveals what your workflow runs consumed.

Checkpoint Reached

Current n8n Cloud Gateway credits balance recorded as the before reading.

Can't See Your Balance?

Confirm that you are using an n8n Cloud instance.
Switch to the instance owner account if Admin Dashboard is missing from the sidebar.
Use an eligible Cloud account that displays a balance before continuing.

Need help locating the balance? Help me find Gateway credits in my n8n Cloud account.

Confirm workflow editor access

The workspace contains the editor where you build workflows. Returning there confirms that your account can move between instance settings and the building area.

Before you switch back, what should the workspace show if your account is ready to build?

Select Workspace in the Cloud admin dashboard.
Check for either Start from Scratch or Create Workflow on the workspace page.

You should see one of those workflow creation options. That confirms your account can access the workflow editor.

That's the setup locked in. Your balance baseline is saved for the final comparison.

Your n8n Cloud account is ready. Next up, you'll create the workflow that sends its first model prompt without attaching a credential.

Test a Gateway-Powered Model

The starting Gateway credits balance you recorded gives this workflow a clear before reading. You can later compare it with the amount left after your test runs.

A first model test often stalls while you create a provider account or API key. Gateway credits let an AI model in n8n Cloud answer a prompt without an attached provider credential.

In this step, get ready to:

Create a new workflow with a supported AI model node.
Configure the model node to use Gateway credits.
Confirm a one-line model reply in the output panel.

Create a workflow and add a model node

A workflow canvas holds the nodes that process your data. This first workflow starts with one model node so you can test the credit path on its own.

Return to the Workflows list from the editor you used earlier.
Click Create Workflow in the upper-right corner.

You should see a blank workflow canvas or a location choice for the new workflow.

Select the same space you are already using if n8n asks where to store the workflow.
Select the add-node control on the blank canvas.

You should see the nodes panel open beside the canvas.

Search for AI model in the nodes panel.
Select a supported AI model node that offers Gateway credits.

You should see the model node on the canvas. Its configuration panel should also be visible.

Can't find a supported model?

Clear any filters in the nodes panel.
Search for AI model again.
Choose a model node whose credential setup includes Gateway credits.

Still stuck? Help me find a Gateway-supported AI model node in my n8n Cloud workflow.

Configure Gateway credits and add a prompt

Credential setup can feel risky because it usually means handling a secret. Here, n8n manages the gateway route so you do not paste or reveal an API key.

Select the model node on the canvas to return to its configuration panel.
Choose Gateway credits in the credential setup.

You should see Gateway credits selected without a provider API-key credential attached.

Enter Reply with one sentence explaining why concise summaries help busy readers. in the prompt field.

You should see the complete one-line prompt ready inside the model node.

How does the gateway handle this request?

The model call uses the shared prepaid balance in your n8n instance. You do not need a separate provider account.

The selected provider receives the content of your prompt. Your identity and n8n account details stay with n8n.

Can't select Gateway credits?

Confirm that the selected node is a supported AI model service.
Return to the nodes panel if Gateway credits is absent.
Select another supported model node that offers Gateway credits.

Need a hand? Help me configure my n8n AI model node to use Gateway credits without a provider API key.

Run the model and inspect its reply

Executing only the model node isolates this first test. A successful reply proves that the node can reach a model through Gateway credits.

Before you run the node, do you expect it to request an API key or return a reply using your Gateway credits?

Click Execute step in the model node panel.
Select the OUTPUT panel when the run finishes.

You should see a reply that explains why concise summaries help busy readers. The exact wording can vary between models.

Read the one-sentence reply in the node output panel.

That is your first Gateway-powered model result. The node answered without an attached provider API-key credential.

No reply in the output panel?

Check that the model node still shows Gateway credits in its credential setup.
Confirm that the prompt field contains the complete one-line prompt.
Run Execute step again after completing the node configuration.

Still no reply? Help me troubleshoot why my Gateway-powered n8n model node is not returning output.

Your model can now answer through Gateway credits. Next up, you'll test a document or web tool service on its own.

Extract Text with a Tool Service

You already proved that your AI model can reply through Gateway credits without a provider credential. A useful summary now needs source text.

A Gateway-powered tool service extracts that source text inside n8n Cloud. You will test the extraction on its own before connecting it to the model.

In this step, get ready to:

Add a supported Gateway-powered document or web tool service without attaching a provider credential.
Configure the tool node with your selected PDF file or page URL.
Confirm a standalone run returns readable extracted text.

Choose a focused source

Extraction is easiest to check when you recognise the source. A short PDF or focused public page gives you a clear result to compare.

PDF file

Select a short PDF on your device that contains visible text.
Use a public or non-sensitive PDF for this practice run.

Page URL

Select a public page with a focused article or documentation section.
Load the page in your browser to confirm it opens without a sign-in.
Record the page address here: https://example.com/article.

What does the provider receive?

Gateway credits send the request content to the selected provider. For this step, that content is the page or document you submit.

The provider does not receive your identity or n8n account details. A public or non-sensitive source keeps this practice run suitable for a screenshot.

Add a Gateway-powered tool node

The tool node is the extraction layer between your source and the AI model. Choose a Gateway-powered service that matches the input you selected.

For web content, you can use Firecrawl. For document workflows, you can use LlamaParse or PDF.co.

Switch back to the n8n workflow editor from earlier.
Click Add step on the canvas.

You should see the node picker while your existing AI model node remains on the canvas.

Search for the supported service that matches your source.
Select the service from the node picker.

You should see the new tool node configuration panel. Its available operations reflect the source type that service can process.

Choose an operation that extracts text from one document or page.
Select Gateway credits instead of adding a provider credential.

You should see source input controls without a provider credential attached.

PDF file

Use the node's file input control to select the PDF from your device.

Page URL

Paste https://example.com/article into the node's URL input.

Return to the canvas using the node panel's close control.

You should now see the AI model node plus the Gateway-powered tool service node. The tool node keeps your selected source in its configuration.

Run the tool node on its own

A standalone run isolates the extraction layer. This gives you one place to inspect before the model is involved.

The next action uses Gateway credits from the balance you recorded earlier. A single node run keeps the test controlled.

Before you run the node, do you expect this stage to return source content or a model-written summary?

Select the tool node on the canvas.
Run only the tool node by clicking Execute step.

You should see readable extracted text from the selected source in the node output panel. The wording should match content from your PDF or page.

Scan the first section of extracted text in the node output panel.
Match one visible phrase against your original source.

That is the extraction layer working: your workflow can now turn a document or page into model-ready text without a provider credential.

Output empty or unreadable?

For a page URL, use a public page that opens without a sign-in.
For a PDF, switch to a file with visible text if the output is empty.
Confirm the tool node still shows Gateway credits.

Still stuck? Help me troubleshoot my Gateway-powered tool node extraction.

✔️ I see extracted text

Your document or web tool node has returned extracted text without an attached provider credential.

ⓧ I'd like to double check my setup

Confirm the AI model node remains configured to use Gateway credits.
Confirm the document or web tool node uses Gateway credits without an attached credential.
Confirm the tool node contains your selected PDF file or page URL.
Confirm the latest standalone execution shows extracted text in the node output panel.

Your workflow can now extract text from a document or page without a provider credential. Next, you'll connect that text to the AI model for a five-line summary.

Generate a Five-Line Summary

The extraction test proved that your source can become usable text inside n8n Cloud. Your AI model has also answered a standalone prompt with Gateway credits.

Those two successes still sit in separate nodes. In this step, you will connect them so the extracted text becomes a five-line summary.

In this step, get ready to:

Route the extracted text from the tool node into the model node.
Replace the test prompt with a five-line summary instruction.
Run the complete workflow and inspect the final summary.

Connect the tool to the model

A connection gives n8n a route for passing data between nodes. Once this route exists, the model receives each item produced by the tool.

Return to the workflow canvas from the tool node output panel.
Select the grey dot or Add node control on the right side of the tool node.
Drag the arrow to the grey rectangle on the left side of the model node.

You should see a connection line running from the tool node to the model node. That line confirms the extracted data has a path into the model.

Connection Going the Wrong Way?

Hover over the misplaced connection.
Select Delete.
Create the connection again from the tool node to the model node.

Still stuck? Help me connect my n8n tool node output to my AI model node.

Map the extracted text into the prompt

Data mapping references a value produced by an earlier node. The mapped field changes automatically whenever the tool extracts text from a new source.

Select the AI model node on the workflow canvas.
Locate the field containing your earlier one-line test prompt.
Replace that test prompt with Summarize the following extracted text in exactly five lines:.
Drag the field containing the extracted text from the INPUT pane into the prompt field below the instruction.
Confirm the prompt preview shows text from your selected PDF file or page URL.
Confirm the model node still uses Gateway credits with no provider credential attached.

The prompt should now contain one fixed instruction followed by the mapped extracted text. Each workflow run can therefore summarize the source that entered through the tool node.

Why Use the UI Mapper?

The mapper inserts a reference that resolves during each workflow run. A different source can flow through the same prompt without manual copying.

Extracted Text Missing From the Prompt?

Run the tool node from earlier again if the input pane contains no data.
Drag only the field containing the extracted text into the prompt.
Confirm the connection line still runs from the tool node to the model node.

Need another pair of eyes? Help me map my extracted text into the n8n model prompt.

Run the full workflow

A standalone node test proves that one service works. A full execution proves that the source can travel through the complete workflow.

This full run uses Gateway credits for its service calls. One deliberate run is enough to confirm the pipeline.

Before you run it, do you expect the model to receive the extracted text automatically?

Click Execute Workflow at the bottom of the workflow editor.

You should see the tool node run before the model node. This order shows that n8n extracts the source text before requesting the summary.

Select the AI model node after the execution finishes.

You should see a five-line summary in the output panel. The content should reflect the PDF file or page URL processed by the tool node.

That is the complete document-to-summary path working. Your source now moves through two Gateway-powered services without an attached provider credential.

Summary Not Looking Right?

Check that the connection line starts at the tool node.
Confirm the model prompt still contains the mapped extracted-text field.
Confirm the instruction still asks for exactly five lines if the response has the wrong length.

Still seeing the wrong result? Help me troubleshoot my n8n five-line summary workflow.

✔️ Awesome, I've got everything!

Excellent. Your workflow now turns the selected PDF file or page URL into a five-line summary without provider credentials.

ⓧ I'd like to double check the workflow

Confirm the tool node still points to your selected PDF file or page URL.
Confirm the tool node still uses Gateway credits without an attached provider credential.
Confirm a connection line runs from the tool node to the AI model node.
Confirm the model prompt includes the mapped extracted text.
Confirm the model prompt requests a five-line summary.
Confirm the model node output shows the successful five-line summary.

Your workflow now produces a summary from a real source. Next, you will inspect the execution record and compare your remaining Gateway credits with the before reading.

Review Usage and Export the Workflow

Your n8n Cloud workflow now sends extracted text into an AI model. The model returns a five-line summary without provider credentials.

The final review makes that result auditable. You will also export a portable copy of the completed workflow.

In this step, get ready to:

Review the successful execution node by node.
Compare the current Gateway credits balance with your before reading.
Download a portable workflow copy with no provider credentials or API keys.

Review the successful execution

An execution record shows how data moved through the successful run. Checking both outputs proves that the tool-to-model handoff worked.

Before you open the run, predict which node shows the extracted text first.

Select Executions in the workflow's top menu.
Select the successful run you completed earlier.
Select the document or web tool node on the execution canvas.
Review the extracted text in its output panel.
Select the AI model node.
Review the five-line summary in its output panel.

You should see extracted text in the document or web tool node's output. You should see the five-line summary in the AI model node's output.

What does the execution log prove?

The execution log preserves the data that passed through each node during the run. Matching outputs confirm that the model received the text produced by the tool node.

You now have a node-by-node record of the successful summary workflow.

Can't find the successful run?

Confirm that you selected Executions from the workflow you ran earlier.
Look for the most recent successful execution in the list.
Return to the execution view after selecting the successful run.

Still stuck? Help me find and inspect my successful n8n workflow execution.

Compare the Gateway credits balances

Each supported service call draws from the shared Gateway credits balance. Comparing the current value with your before reading reveals the usage recorded between those checks.

Before you check the balance, predict whether the value has changed since the successful run.

Return to the n8n Cloud account settings page from earlier.
Locate the current Gateway credits balance.
Record the current value here: your current Gateway credits balance.
Compare your current balance with the before reading you recorded earlier.
Calculate the difference between the two readings.

The difference shows how many Gateway credits the workflow activity consumed between the two checks.

You now have a real usage figure for a successful summary run.

Can't see the current balance?

Confirm that you returned to the same Cloud account settings page where you recorded the before reading.
Refresh the settings page to load the current balance.
Confirm that you are viewing account settings instead of workflow settings.

Need a hand? Help me locate and compare my n8n Gateway credits balances.

Export and inspect the workflow

A JSON export is a portable definition of your workflow. Because neither node has a provider credential attached, this copy should contain no provider API key.

Before you download the workflow, predict whether the exported file contains any provider credentials.

Return to the workflow editor from earlier.
Select the three dots in the upper-right corner of the top navigation bar.
Select Download.

Your browser should download the current workflow as a .json file.

Open your browser's downloads panel.
Open the downloaded .json file.
Inspect the entries for the document or web tool node.
Inspect the entry for the AI model node.
Confirm that neither node contains an attached provider credential or API key.

You should find both configured nodes with their connection preserved. You should find no attached provider credentials or API keys.

What is included in the export?

n8n saves workflows in JSON format. An export can include credential names and IDs when a workflow uses them.

Your Gateway-powered nodes have no provider credentials attached. You can import this copy into another n8n instance without carrying provider secrets.

Download missing or credentials present?

Open your browser's downloads panel if the file is not visible.
Confirm that the workflow download finished.
Return to the workflow editor if the export contains provider credential references.
Check that both supported nodes still use Gateway credits.
Download the workflow again after removing any attached provider credential.

Need help checking the export? Help me verify that my n8n workflow JSON contains no provider credentials or API keys.

Your run is now traceable through the execution log. The portable workflow copy is ready to keep or share without provider secrets.

Secret mission

Audit a Second Summary for Accuracy

Run your workflow against a second source. Trace every summary line back to the extracted text. You will finish with evidence that the automation is repeatable.

Clean Up Your Resources

Your n8n Cloud workflow can stay active, remain saved while inactive, or be removed. Supported node calls consume Gateway credits.

Cost warning

Each supported service call uses Gateway credits. Scheduled runs can continue reducing your balance after you leave the editor.

Choose the Pause or Delete option if you do not plan to run the workflow again soon.

Resources you used:

The completed n8n Cloud workflow containing a Gateway-powered document or web tool node connected to an AI model node.
The exported workflow JSON file saved on your computer.

Keep everything running

No action needed. This option suits active testing when you still want to run documents or page URLs through the workflow.

Keep the workflow active only while you are testing it.
Check your Gateway credits balance after each test run.
Keep the exported workflow JSON file as a portable backup.

Pause - I'll come back to this later

This option stops future scheduled runs while preserving the workflow for another session.

Switch back to the workflow editor from earlier.
Deactivate the completed workflow.

You'll see the workflow status change to inactive. That stops future scheduled runs from consuming more Gateway credits.

Keep the exported workflow JSON file in its current saved location.

Your workflow is now parked for later. The exported JSON remains available as a portable backup.

Delete - I don't want to use this again

Deleting both copies is final, so use this option only when you no longer need the workflow or its portable backup.

Return to the workflow editor from earlier.
Deactivate the completed workflow.

You'll see the workflow status change to inactive. This prevents another scheduled run while you remove it.

Use the workflow's delete option to remove the completed workflow.
Return to the workflow list.

You should no longer see the completed workflow in the list. That clears the hosted copy from your account.

Use Finder on macOS or File Explorer on Windows to return to the exported workflow JSON file's current saved location.
Delete the exported workflow JSON file.

You should no longer see the exported file in that location. Your project copies are now removed.

Nice Work!

You did it! You built an n8n Cloud workflow that turns text from a PDF file or page URL into a five-line AI summary without provider API keys.

You've learned how to:

Use Gateway credits to run supported document or web services. You also used them to run an AI model without attaching provider credentials.
Connect extracted text to an AI model. Your completed workflow now produces a five-line summary from a selected PDF file or page URL.
Review the execution log to trace each node's output. You compared the Gateway credit balances to measure consumption. You also exported a portable JSON workflow with no provider credentials or API keys.
Complete the optional Secret Mission to extend the workflow through an additional challenge.

Ready to quiz yourself?