A service invoice that came as a PDF only, scanned pages, a photo of a receipt, a till slip, a contract. IngestCore watches a folder, reads with OCR and AI running on your own computer — and returns structured data, checked and ready to integrate.
Nobody opens invoices one by one anymore. Almost every company already has some RPA doing that. The problem is not the lack of automation: it is what the automation does not cover — and the manual work that comes back because of it.
The bottleneck is not reading the easy document — it is what is left when the current automation gives up.
Here everything comes in through the same line: no XML is not a blocker, and whatever cannot be read shows up as a pending item instead of vanishing.
Seven steps between the document that arrives and the data your system consumes. OCR is only one of them.
Watched folder, e-mail or API. The document arrives without anyone typing.
Native XML when it exists — exact and free. OCR and visual reading when it is paper or a photo.
Deterministic analysis before any model: invisible text, instructions aimed at the AI, encoded content.
Sensitivity and document type decide whether reading runs locally, in the cloud, or not at all.
Rules check item totals, tax ID check digits and date consistency. Below the threshold it goes to human review.
Twenty input formats become a single contract, with the same field names.
Canonical JSON on the way out, ready for the ERP, API or automation adapter.
The installer ships OCR, the PDF reader and the local AI manager inside. On first run, IngestCore analyses your computer and recommends the best AI setup — you just click.
Integrated Tesseract, Portuguese and English, multi-pass with automatic selection of the best reading.
Analyses CPU, memory and GPU and recommends the right model for your machine — textual and visual reading.
Deterministic analysis against malicious content, automatic quarantine and a confidentiality matrix per document type.
Invoices, receipts, bank slips, contracts — each type with its own pipeline, fields and review rules.
Canonical JSON, CSV and integration layouts; quarantined documents never leave.
Cloud AI providers are optional, BYOK, with policy, budget and per-type confidentiality — blocked by default.
Whoever writes the integration does not care whether the document was XML, a photo or a scanned PDF. They want the same object, with the same fields, every time. That is what IngestCore delivers — and it is what lets you write the adapter once.
For whoever integrates
e-invoices · bank slips · transport documents · receipts · contracts · payment proofs · images · scans
JSON
issuer · recipient · amounts · line items · dates · key · confidence · source
Factory default. OCR and models run on the machine; the document never crosses the internet.
Only when there is a connected provider, a policy that allows it and a budget above zero.
Local first; the cloud steps in only when local confidence falls below the defined threshold.
Not every document needs — or should — leave the company. Policy decides, not code.
IngestCore processes everything on 127.0.0.1: local database, local OCR, local AI. No document is sent to an external service without three explicit decisions of yours — a connected provider, a policy that allows it and a budget above zero. Your IT team can verify it with their own firewall.
Drag the controls to match your operation.
Basis: ~3 minutes per document to open, read, type and check.
Processing runs on your machine — that is why local OCR and AI are unlimited on every plan. The license is per computer, activated offline.
Need IngestCore shaped to your workflow? Custom document types, export layouts and integrations at R$ 150/hour.
1,000 documents or 120 days, whichever comes first. No card field anywhere.
No credit card, no sign-up and no sales call. Install it, point it at a folder and see the result on your own invoices. If the 1,000 documents run out, your data stays with you — always.
The cases below use document types that ship with the product — nothing has to be developed to start.
Electronic invoices in XML and their printed counterparts in PDF: issuer, line items, taxes and access key extracted and checked before they become entries.
Bank slips and payment proofs read from a folder or e-mail, with amount, due date and payment line ready for reconciliation.
Freight documents, for whoever needs to tie shipping to the tax document it travels with.
Receipts photographed on a phone — the case where local visual reading makes the difference.
Parties, dates and key clauses extracted, with human review wherever confidence is low.
Scanned paper in bulk, with multipass OCR and the same output contract as digital documents.
The product transforms the document. XCODEF connects the result to your business: we are the engineering team that built IngestCore, and we deliver services on top of it.
Talk about integrationNo. OCR, AI and the database run locally. A cloud service only works if you connect your own provider (BYOK), allow it in the policy and set a budget — and even then, confidential types and quarantined documents never leave.
Ingestion of new documents pauses. Everything already processed stays accessible, searchable and exportable — your data is never held hostage. Activate the license and ingestion resumes where it stopped.
No. The installer brings everything. For local AI, IngestCore analyses your computer and installs the recommended model with one click — or you can use OCR only.
Yes. The whole product works offline, including license activation. Internet is only used if you choose cloud AI or download models.
Per computer, valid for one year. At purchase you provide the machine code (shown in the product) and receive the key by e-mail — paste it, activate. Changed machines? XCODEF reissues it.
IngestCore exports canonical JSON and CSV; specific ERP layouts are built to order (R$ 150/h) and become part of your installation.