Resources · n8n integration

Mistral AI n8n integrationAutomate Mistral AI with n8n.

Scanned paperwork holds data nobody wants to retype. The Mistral AI n8n integration exposes 1 resource and 1 operation, Extract Text, which reads a document or an image with OCR and hands the text to the next node. No trigger ships with it, so a schedule or another app starts the run.

Try the Mistral AI integration on n8nSee the 1 operations

Verified Trustpilot reviews · AI, automation & growth agency

Why automate

What does the Mistral AI n8n integration actually do?

The Mistral AI n8n integration wires a workflow to Mistral's OCR API through a single node. One resource, Document. One operation, Extract Text. You point it at a PDF or an image, and it returns the readable text of that file inside the workflow. OCR stands for optical character recognition, the step that turns a scan into characters a workflow can search, parse or store.

Take supplier invoices piling up in a shared folder. A schedule picks up the new files, the Mistral AI node reads each one, and a parsing step writes the amounts into Google Sheets. The accounting export stops being a retyping job and becomes a row that already exists when someone opens the file.

Expense receipts work the same way, with a shorter loop. A photo of a receipt reaches the workflow as binary data, Extract Text pulls the merchant and the total out of the image, and a message in Slack tells the person who submitted it that the claim went through. Nobody opens an attachment to read four numbers.

Recruiting is the third obvious one. CVs arrive as PDFs, the node turns each one into plain text, and the text feeds whatever scoring or sorting logic you already run before a reply goes out through Gmail. Same for contracts and reports that need to be searchable rather than filed away as images.

The node stops at OCR. Mistral's API does more than read documents, and none of the rest is exposed here: chat completions, for example, live outside this node. When you need them, the HTTP Request node calls any endpoint of the API and reuses the same Mistral Cloud credential through predefined authentication, so you are not managing a second key. Extracted text that has to change language goes to a dedicated node such as DeepL rather than back to Mistral.

Two limits are worth knowing before you build. A file sent as binary data stays under 50 MB, which n8n reminds you in the field itself, and the model is fixed: mistral-ocr-latest is the only choice offered. There is also no Mistral AI trigger, so the workflow always starts somewhere else. If you are still deciding which platform runs this kind of pipeline, the n8n review covers hosting and cost in detail.

Connect

How do you connect Mistral AI to n8n?

  1. 01

    Create the API key in Mistral

    Sign in to your Mistral account and open the API Keys page of the console. Before a key does anything, payment information has to be added in Workspace > Billing and payments activated on the account: Mistral only issues working keys to accounts set up for billing. Select Create new key, then copy the value straight away and keep it somewhere safe for the next step.

  2. 02

    Store the Mistral Cloud credential in n8n

    In n8n, open the Credentials menu and create a Mistral Cloud credential. It asks for one thing, the API Key you just copied. A credential is saved once and reused by every workflow on the instance, and this one also authenticates the Mistral Cloud chat model and the Embeddings Mistral Cloud sub-nodes, so a single key covers your OCR and your language model calls.

  3. 03

    Run a first document through the node

    Drop a Mistral AI node into a workflow and pick the credential from its dropdown list. Leave the resource on Document and the operation on Extract Text, set Document Type to Document for a PDF or Image for a photo, then choose how the file arrives with Input Type. Execute the node on one real file and read the output before you build anything around it.

Actions

What the Mistral AI node can do

The Mistral AI node exposes 1 operations across 1 resources. For each one: the node as you configure it in n8n, the required fields, and our field notes.

Resources × operations matrix
ResourceExtract Text
Document✓

Document

1 operation
01

Extract text

document.extractText

What you see in n8n

Notes & use cases

Extract Text sends a file to Mistral's OCR engine and returns its text to the next node.

Key parameters

  • Model: mistral-ocr-latest, the only value offered.
  • Document Type: Document for a PDF, Image for a photo or a scan.
  • Input Type: Binary Data reads a file already carried by the workflow, URL lets Mistral fetch it.
  • Input Binary Field: the name of that binary field, data in most workflows, under 50 MB and 1,000 pages.
  • URL: the file address, often {{ $json.field }} from the node before.
  • Enable Batch Processing: several documents per call, sized by Batch Size.
Use cases
making signed contracts searchable, or reading a photographed receipt so the amount reaches the expense sheet untouched.
Need help

Need help automating Mistral AI with n8n?

A person reads every message.

FAQ

Mistral AI and n8n: common questions

01Is the Mistral AI n8n integration free?
Yes, on the n8n side. The node ships with n8n, so there is nothing to install and no add-on to buy, whether you work on n8n Cloud, the hosted offer from n8n, or on your own instance installed with Docker or npm. A workflow is identical in both cases. The Mistral side is where money enters: Mistral asks you to add payment information in Workspace > Billing and to activate payments before an API key returns anything, so the OCR calls run on a billable Mistral account. n8n itself never charges for the node or for the credential.
02What credentials does the Mistral AI node need?
One credential, Mistral Cloud, authenticated with an API key. You create it once in the Credentials menu, paste the API Key copied from the API Keys page of the Mistral console, and every workflow picks it from the node dropdown. The node has no authentication selector of its own: the API key is the only supported method. The same credential also covers the Mistral Cloud chat model and the Embeddings Mistral Cloud sub-nodes, which is useful when an OCR step and a language model step sit in the same workflow and you would rather manage one key than three.
03What are the limits of the Mistral AI node in n8n?
It depends what you expect from it. The node exposes 1 resource, Document, and 1 operation, Extract Text, with a single model value, mistral-ocr-latest. A file passed as binary data stays under 50 MB, a limit n8n prints in the field hint along with a cap on the number of pages. Everything else in the Mistral API sits outside the node, chat completions included: the HTTP Request node calls those endpoints and reuses the same Mistral Cloud credential through predefined authentication. When several files go out together, the node options offer batch processing and the deletion of files from Mistral Cloud once they are processed.
04Does Mistral AI have a trigger node in n8n?
No. There is no Mistral AI trigger, so a workflow that runs OCR starts somewhere else. Three routes cover almost every case: the Schedule Trigger at a fixed interval, for instance a morning pass over a folder of new files; the n8n Webhook node, which is a URL a third party service calls when something happens on its side; or the trigger of another app already connected to the workflow, such as a file storage or a mailbox. Once the run starts, the document reaches the Mistral AI node as binary data or as a URL and the extracted text moves on.
05n8n or Make for Mistral AI?
Both run this kind of OCR step, and the decision sits elsewhere. Make is hosted by Make, with no self-hosting option, and billed per operation, which is easy to forecast on small volumes. n8n runs on your own server or on n8n Cloud, and that matters when the files going through OCR are contracts, payslips or health records you would rather keep on infrastructure you control. The cost model differs in the same way: per operation on one side, per instance or plan on the other. Visual logic differs too, so try one document on each before committing.
Same platform

Other docs you might need on this stack.

Browse all integrations