Last updated: 2026-08-16
Mindee is a document-processing API founded in 2018 that converts invoices, receipts, passports and IDs into structured JSON. Its REST API and SDKs for Python, Node.js, Java, PHP, Ruby and .NET offer pretrained models for common documents plus custom model training, backed by SOC 2 Type II certification and EU/US data residency.
About Mindee
Mindee is a document-processing API founded in Paris in 2018 that gives developers a single REST API to turn unstructured documents into structured JSON. It is built around three pillars: a self-service platform where builders train their own extraction models, a catalog of pretrained models for common document types (invoices, receipts, passports, driver's licenses, bank statements, resumes, contracts), and a custom-model service for large customers with document types the pretrained catalog does not cover. The API combines OCR, computer vision and machine learning to extract fields, return a confidence score per field, and return bounding-box or polygon coordinates so results can be validated programmatically or routed to a human reviewer. Five model families cover Extraction, Classification, Crop and Split (breaking a multi-document scan into separate files) and OCR, with chaining that runs pre-processing and extraction in a single API call. A Continuous Learning feature lets models improve automatically from new data without a manual retraining cycle. The platform is aimed at backend and platform engineers building document ingestion pipelines, and at product teams in accounting software, accounts-payable automation, expense management and KYC/customer-onboarding who need a reliable data-capture layer rather than a finished front-end product. Official SDKs cover Python, Node.js, Java, PHP, Ruby and .NET, all MIT-licensed and hosted on GitHub, alongside webhook and polling integration patterns and a Zapier connector. Pricing is credit-based: a physical page submitted to the API consumes one credit regardless of document type or format, so a five-page contract costs five credits. A Starter subscription starts at $29/month ($312 billed annually) and a Pro subscription at $119/month ($1,284 billed annually), each combined with a separately purchased credit pack that starts at $30/month for 500 pages and drops toward $0.042 per page at higher volumes; Enterprise is custom-priced and requires an annual contract of 500,000+ credits. A 14-day free trial gives full API access with no credit card required. Mindee has raised $23.4M across four rounds, including a $14M Series A in October 2021 led by GGV Capital with Y Combinator, Alven, Serena Capital and Bpifrance participating. The company holds SOC 2 Type II certification and is GDPR-compliant, and lets customers choose EU or US data residency at the model or organization level.
Pricing
14-day free trial with full API access, no credit card required. Starter from $29/month ($312/year billed annually) plus a separately purchased credit pack starting at $30/month for 500 pages (or $324/year for 6,000 pages on annual billing), with the per-page rate dropping from about $0.06 toward $0.042 at higher volumes. Pro from $119/month ($1,284/year annually) adds live chat support, unlimited seats and data-processing localization, on the same credit-pack pricing. Enterprise is custom-priced, requires an annual contract of 500,000+ credits/year, and adds a dedicated account manager and custom SLAs. Credits are billed per physical page submitted, not per document.
Key Features
- Pretrained document models: Ready-to-call models cover 7+ common business document types out of the box, so most teams skip the training step entirely.
- Custom model training: A no-code trainer builds extraction models for document types outside the pretrained catalog, used by large customers with bespoke formats.
- Confidence scores and bounding boxes: Every extracted field returns a confidence score plus polygon coordinates, so low-confidence fields can be routed to human review instead of auto-accepted.
- Split, Crop and Classify: Split breaks a multi-page upload into separate documents, Crop isolates several documents scanned on one page, and Classify sorts incoming files by type before extraction.
- Chaining: Pre-processing and extraction run in a single API call, cutting a multi-step pipeline down to one request.
- Data residency selection: Pro and Enterprise customers choose EU or US data processing and storage at the model or organization level for GDPR and regional compliance.
- Six official SDKs: MIT-licensed client libraries across six languages are published on GitHub, plus webhook and polling integration patterns for async jobs.
Pros
- Pretrained models for 7+ common document types (invoices, receipts, passports, IDs, bank statements, resumes, contracts) skip the training step entirely, unlike fully custom OCR builds.
- SOC 2 Type II certified with selectable EU or US data residency, covering compliance requirements that many smaller OCR APIs do not offer.
- Confidence scores and bounding-box coordinates on every field let teams route only low-confidence extractions to human review instead of manually checking every document.
- Six official MIT-licensed SDKs (Python, Node.js, Java, PHP, Ruby, .NET) cut integration time versus building against the raw REST API alone.
Cons
- Credits are billed per physical page submitted, so a 5-page contract costs 5x a single-page receipt regardless of extraction complexity, and users on review sites specifically flag this as a hidden cost driver.
- OCR accuracy fluctuates on unclear scans and unusual formats, and reviewers report occasional typos in non-English languages including no dedicated German-language support.
- Data-processing localization (EU/US residency choice) and live chat support are gated to the Pro tier and above, not included in Starter.
- No ongoing free tier, only a time-limited trial; production use requires a paid Starter or Pro subscription plus a separate credit pack purchase.
Frequently Asked Questions
How much does Mindee cost in 2026?
Mindee's Starter plan is $29/month ($312/year billed annually) and Pro is $119/month ($1,284/year annually), each combined with a separately purchased credit pack starting at $30/month for 500 pages, with per-page rates dropping toward $0.042 at higher volumes. Enterprise is custom-priced and requires an annual contract of 500,000+ credits/year. Credits are billed per physical page submitted, so a 5-page document costs 5 credits.
Is Mindee free to use?
Mindee has no ongoing free tier, only a 14-day free trial with full API access and no credit card required. After the trial, production use requires a paid Starter or Pro subscription plus a purchased credit pack, since every plan bills separately for the pages processed.
What are the best alternatives to Mindee?
Reducto is a closer fit if the extracted data feeds a RAG pipeline or AI agent rather than a business workflow. LlamaIndex suits teams already building a broader document-and-retrieval orchestration layer who want extraction as one step in a larger agentic pipeline. AWS Textract or Google Document AI are options for teams already committed to that cloud provider's ecosystem.
How does Mindee compare to Reducto in 2026?
Mindee leads with pretrained, ready-to-call models for common business documents like invoices, receipts and IDs, so most teams skip training entirely. Reducto is built for feeding structured document data into RAG pipelines and AI agents across 30+ file types with agentic OCR correction. Choose Mindee for a defined document-extraction workflow, Reducto if the output is going straight into an LLM pipeline.
How do you get started with Mindee?
Sign up for the 14-day free trial at mindee.com with no credit card required, then call a pretrained model or train a custom one in the self-service console. Integrate with the REST API directly or an official SDK for Python, Node.js, Java, PHP, Ruby or .NET, and use the returned confidence scores to route low-confidence fields to human review.
Top Alternatives
- Reducto: Pick Reducto if your pipeline feeds RAG or AI agents; pick Mindee if you need pretrained invoice, receipt and ID extraction out of the box.
- LlamaIndex: Pick LlamaIndex if you're building a broader agentic document-and-retrieval workflow; pick Mindee for a dedicated, pretrained document-extraction API.