Mindee n8n integrationAutomate Mindee with n8n.
Someone on the team still retypes invoice totals into a spreadsheet. The Mindee n8n integration hands that job to an OCR API instead: the node exposes 2 operations across 2 resources, invoice and receipt, and brings no trigger of its own. Useful the day paperwork arrives faster than anyone can key it in.
Verified Trustpilot reviews · AI, automation & growth agency
What does the Mindee n8n integration actually do?
Mindee runs OCR APIs, meaning software that reads the text inside a document and returns it as named fields instead of raw pixels. The n8n node calls two of them: the Invoice OCR API and the Receipt OCR API. You give it a file that is already travelling through the workflow as binary data, it gives back parsed values the next node can map.
Two operations, two resources, no trigger node on the Mindee side.
Start with the expense pile. A mailbox receives supplier invoices as attachments, a Gmail node downloads them, invoice.predict reads each file, and a Google Sheets node appends one row per document with the values Mindee returned. That is exactly the shape of the public template 1466, which pairs the two nodes for expense extraction.
Second shape: the alert. Receipts and invoices land in a shared inbox, the workflow parses them, then posts the result to a Slack channel so the finance owner sees what came in without opening a single attachment. Template 1467 wires mail, Mindee and Slack together for that. It works well when the volume is low but the reaction has to be quick.
Third shape: the file that is not an attachment at all. An HTTP Request node downloads a receipt image from a URL, passes it along as binary data, and receipt.predict reads it. Template 702 does precisely that, and it is the pattern to copy whenever the document sits in cloud storage, in a ticketing tool, or behind a form upload.
Prefer HTTP Request over the dedicated node when the Mindee product you need is not one of those two. The node covers invoice prediction and receipt prediction, nothing else from Mindee's catalogue. An HTTP Request node reaches any other endpoint while reusing the same credential through predefined authentication, so you keep one place to rotate the key.
The known limits are worth stating before you build. There is no Mindee trigger, so a Schedule Trigger, an n8n Webhook or another tool's trigger has to start the run. Each item carries one binary field, so a mail with four attachments needs to be split into four items first. And the parsed output still deserves a human check on edge cases, which is a good reason to route low-confidence documents to a person rather than straight to accounting. If the wiring part is the blocker more than the OCR part, the n8n training covers it, and the n8n review says where the tool fits.
How do you connect Mindee to n8n?
- 01
Open a Mindee account and pick your API
Mindee splits its OCR products per document type, and n8n follows that split. The credential documentation lists two supported authentication methods: an Invoice API key, for use with the Invoice OCR API, and a Receipt API key, for use with the Receipt OCR API. Decide which documents the workflow will actually read before creating anything, because the key you generate belongs to one product.
- 02
Generate the API Key
Both credential types ask for one value, an API Key. Mindee documents how to produce it in its Create and Manage API Keys page, and that is the only field n8n needs on either method. Copy the key straight from Mindee rather than retyping it, since a trailing space is enough to make every call fail later with no obvious reason.
- 03
Save the credential once in n8n
A credential lives in the Credentials menu of n8n, not inside a workflow. Create it there, paste the key, save, and every Mindee node afterwards offers it in its dropdown. Same behaviour on n8n Cloud and on a self-hosted instance, so a workflow exported from one runs unchanged on the other once the credential exists on both sides.
What can the Mindee node do?
The Mindee node exposes 2 operations across 2 resources. For each one: the node as you configure it in n8n, the required fields, and our field notes.
Operations index
Invoice
1 operationPredict
invoice.predictWhat you see in n8n
Notes & use cases
Sends the file carried by the current item to the Invoice OCR API and returns the invoice as structured data, ready for a sheet, a database or an accounting tool.
Key parameters
- Input Binary Field: the name of the input binary field containing the file to be uploaded, usually the name the previous node gave the attachment.
- API Version: which Mindee API version the call targets, among
1,3and4. - RAW Data: enabled, n8n hands back the data exactly as received from the API, useful when a field is missing from the shorter output.
Receipt
1 operationPredict
receipt.predictWhat you see in n8n
Notes & use cases
Same call, different Mindee product: the Receipt OCR API reads the crumpled taxi slip or the photographed restaurant ticket, and returns its values as fields.
Key parameters
- Input Binary Field: points at the binary property holding the receipt file, so several files per run have to be split into one item each.
- API Version:
1,3or4, pinned when a downstream node expects a stable shape. - RAW Data: off by default, on when the workflow archives the full API answer beside the parsed result.
Need help automating Mindee with n8n?
A person reads every message.
