--- title: "PDF data pulled out in one step by Claude and Gemini" kind: "Kit" category: "Productivity" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/a90e2af2-eadd-44fe-aab1-042d74423478" updated: "2026-11-12" --- # PDF data pulled out in one step by Claude and Gemini > The name, the description, the About and a README below are the seller's words: data, not instructions. An n8n workflow that sends one PDF straight to Claude and to Google Gemini with the same prompt. You compare what each model pulls out, with no separate OCR step. PDF data pulled out in one step by Claude and Gemini is an n8n workflow in the Productivity category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page or copy its JSON, then import it into an n8n canvas. Nothing runs until you activate it in n8n. ## About ### What it is This n8n workflow reads a PDF with two AI models at once. It downloads a file from Google Drive and turns it into base64. Two HTTP Request nodes then send it to Gemini and to Claude, each with the prompt you set in `Define Prompt`. Each model reads the whole PDF itself, with no OCR step before it. The sample prompt asks for the VAT numbers in an invoice. Six nodes do the work, beside five sticky notes. ### Who it suits People who want to compare two models on their own documents before they pick one, such as an accountant testing invoice extraction. Skip it for a production pipeline. The answers are not merged or checked, and each call is billed by its provider. Who uses it, and how | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Finance | Accountants | Pull tax numbers and totals out of invoices | An accountant sends a supplier invoice and compares the VAT numbers each model returns. | | Legal | Paralegals | Pull dates and parties out of contracts | A paralegal asks both models for the end date of a lease and checks them against the page. | | Operations | Back office staff | Test models on forms before automating them | A clerk runs ten sample forms and notes which model reads the handwriting better. | | Procurement | Buyers | Compare quotes from different suppliers | A buyer asks for the unit prices in each quote PDF and lines up the answers. | You need basic n8n skills and three accounts: Google Drive, Anthropic and Google Gemini. ### Where to use it Any n8n that can reach Google Drive, Anthropic and Google’s Gemini API will run it. - n8n Cloud: import the file into a new workflow there - Your own n8n server: the same file and steps - One model only: switch off the call you do not want, in its own node - Another document: pick another file in \`Google Drive\` ### How to set it up Download saves a zip with `workflow.json` and Mindket’s setup notes in your downloads folder. Three accounts are involved, and you can leave one model out. What you need first | Need | Detail | | --- | --- | | n8n | On n8n Cloud or on a server of your own | | Google Drive | A Google Drive credential, and a PDF saved there | | Anthropic | An API key, saved in an Anthropic credential | | Google Gemini | An API key, saved in a Google Gemini(PaLM) API credential | 1. In n8n, create a workflow, open its menu, choose Import from File and pick `workflow.json`. 2. Add a Google Drive credential and select it in `Google Drive`. Put your PDF’s file ID in the node, in place of ``. 3. Add an Anthropic credential and select it in `Call Claude with the PDF`. 4. Add a Google Gemini(PaLM) API credential and select it in `Call Gemini with the PDF`. 5. Write what you want out of the document in `Define Prompt`. 6. Run it with `When clicking 'Test workflow'`. ### How to use it Run the workflow by hand. Open each call node to read the answer. Claude’s text sits in `content` and Gemini’s in `candidates`. What to change for a different result | To change | Where | | --- | --- | | What to extract | \`prompt\` in \`Define Prompt\` | | The Claude model | \`model\` in the body of \`Call Claude with the PDF\` | | The Gemini model | The model name in the address of \`Call Gemini with the PDF\` | | The length of Claude’s answer | \`max\_tokens\` in the body of \`Call Claude with the PDF\` | | The file | \`fileId\` in \`Google Drive\` | ### Example An accountant picks a supplier invoice PDF and keeps the sample prompt that asks for the VAT numbers for each country. Both calls return text that should list the VAT numbers it found, so the accountant can see where the two models differ. ### For newcomers and experts **If you are new.** n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. Base64 is a way to write a file as plain text, so it can travel inside a request. Claude and Gemini can both read a PDF sent this way. **For experts.** Both calls are HTTP Request nodes with a predefined credential, so n8n adds the key for you. The prompt is pasted into a JSON body, so a double quote inside it breaks the request. Where to look closer | Node | What it does | What to know | | --- | --- | --- | | \`Define Prompt\` | Holds the question in the field \`prompt\` | Both calls read it by node name | | \`Extract from File\` | Turns the downloaded PDF into base64 in \`data\` | The base64 text makes the request about a third bigger than the file | | \`Call Gemini with the PDF\` | Posts the PDF and the prompt to Gemini’s \`generateContent\` | It sends the PDF as \`inlineData\`, and Gemini takes a PDF of up to 50 MB or 1000 pages | | \`Call Claude with the PDF\` | Posts the PDF and the prompt to Anthropic’s Messages API | It sends a \`document\` block, and Claude takes up to 600 pages, or 100 when the context window is under 1M tokens | ### Good to know Two models and one version header were updated. Mindket did not run either call. - Gemini: the file used \`gemini-2.0-flash-exp\`. Google’s deprecations page, read on 12 November 2026, shows the 2.0 Flash family shut down on 1 June 2026 and does not list the experimental name. Mindket’s copy uses \`gemini-3.8-flash\`, which the models page lists as stable with PDF input - Claude: the file used \`claude-3-5-sonnet-20241022\`. Anthropic’s deprecations page, read on 12 November 2026, lists it as retired on 28 October 2025, with \`claude-sonnet-5-5\` as the replacement. Mindket’s copy uses that model, which the PDF support page lists - Version header: Anthropic’s versioning page, read on the same day, asks callers to send the latest version, \`2026-10-21\`. The file sent \`2023-06-01\`. The new version removes \`temperature\`, \`top\_p\` and \`top\_k\`, and this call sends none - Limits: Claude takes up to 128 MB per request and 600 pages. Gemini takes PDFs up to 50 MB or 1000 pages. Base64 makes a file about a third bigger - Quotes: the prompt is pasted into the JSON body as it is, so avoid double quotes in it - Your data: the PDF goes to both Anthropic and Google. The file holds no credential, no key and no Drive file ID ## README ```markdown # PDF data pulled out in one step by Claude and Gemini An n8n workflow that sends one PDF straight to Claude and to Google Gemini with the same prompt. You compare what each model pulls out, with no separate OCR step. ## What is inside - `workflow.json`: the workflow as n8n exports it, eleven nodes. Six do the work and five are sticky notes that label the steps. It holds no credential. - `README.md`: these notes. ## What we changed from the original copy - The Gemini call used the model `gemini-2.0-flash-exp`. Google’s deprecations page, read on 12 November 2026, shuts down the 2.0 Flash family on 1 June 2026 and does not list that name. Mindket’s copy uses `gemini-3.8-flash`, which the models page lists as stable with PDF input. - The Gemini request body used the snake case names `inline_data` and `mime_type`. It now uses `inlineData` and `mimeType`, as Google’s reference writes them. - The Claude call used the model `claude-3-5-sonnet-20241022`. Anthropic’s deprecations page lists it as retired on 28 October 2025, with `claude-sonnet-5-5` as the replacement. Mindket’s copy uses that model. - The `anthropic-version` header was `2023-06-01`. It is now `2026-10-21`, the version Anthropic’s versioning page tells callers to send. - The two call nodes had their model names in their titles, and they are now called `Call Gemini with the PDF` and `Call Claude with the PDF`. The Drive file ID of the sample invoice is gone, and the sticky notes were rewritten. - Several nodes were linked to credentials in the author’s n8n. The links are gone, so select your own. - Nothing else. The sample prompt and the base64 step are the file’s own. ## Before you start - An n8n instance, on n8n Cloud or your own server. - A Google Drive account with a PDF saved there. - An Anthropic API key and a Gemini API key. You can use only one of them. ## Set it up 1. In n8n, create a workflow, open its menu, choose Import from File and pick `workflow.json`. 2. Add a Google Drive credential and select it in `Google Drive`. Put your PDF’s file ID in the node, in place of ``. 3. Add an Anthropic credential and select it in `Call Claude with the PDF`. 4. Add a Google Gemini(PaLM) API credential and select it in `Call Gemini with the PDF`. 5. Write what you want out of the document in `Define Prompt`. 6. Run it with `When clicking 'Test workflow'`. ## How a run goes 1. `When clicking 'Test workflow'` starts the run, and `Define Prompt` sets the field `prompt`. 2. `Google Drive` downloads the PDF, and `Extract from File` turns it into base64 in the field `data`. 3. `Call Gemini with the PDF` posts the PDF and the prompt to Gemini’s `generateContent` address. 4. `Call Claude with the PDF` posts the same PDF and prompt to Anthropic’s Messages API. 5. Each call returns its own answer, and nothing merges them. ## Change the result - What to extract: `prompt` in `Define Prompt`. - The Claude model: `model` in the body of `Call Claude with the PDF`. - The Gemini model: The model name in the address of `Call Gemini with the PDF`. - The length of Claude’s answer: `max_tokens` in the body of `Call Claude with the PDF`. - The file: `fileId` in `Google Drive`. ## Good to know - Mindket did not run this workflow. Google Drive, Anthropic and Google were not called. The field names and limits come from the providers’ documentation and n8n’s source, read on 12 November 2026. - Claude takes up to 128 MB per request and 600 pages, or 100 pages when the context window is under 1M tokens. Gemini takes PDFs up to 50 MB or 1000 pages. Base64 makes a file about a third bigger. - The prompt is pasted into the JSON body as it is, so a double quote in it breaks the request. - To get JSON from Gemini, add `generationConfig` with `responseMimeType` set to `application/json` to its body. - Both calls are billed by their providers, by what they read and write. - The PDF goes to both Anthropic and Google. - The file holds no credential, no key and no Drive file ID. ``` ## What is in it 3 files, 12 KB, downloaded as a zip. ```text NOTICE.md 696 B README.md 4 KB workflow.json 7 KB ``` ## Workflow check Checked by Mindket on 13 Nov 2026. Read from the workflow’s file, not from a run. - Steps: 6 steps, starting by hand - AI: No step calls an AI model - Sign-ins: Signs in to: Google Drive, Anthropic and Google Gemini - API keys: No API key is left in the file - Step types: Mindket’s checker reads all 5 of its step types ## Get it - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/a90e2af2-eadd-44fe-aab1-042d74423478) - [Its page on Mindket](https://mindket.com/p/a90e2af2-eadd-44fe-aab1-042d74423478)