Embedded data import for SaaS products
Mildport turns CSV, Excel, PDF, and other customer files into ready-to-use data inside your product.
Use Mildport in the cloud or deploy a self-hosted version when your data needs to stay within your infrastructure. Fast for developers. Clear for users. Fully under your control.
Every new customer brings a new import challenge.
Then come different formats, unexpected fields, PDFs, mapping issues, and edge cases that turn data import into a project of its own.
“It's just a CSV,” the customer said.
Inside were different column names, unusual date formats, duplicates, and even scanned PDFs.
— Every onboarding team runs into this sooner or later.
Mildport helps turn real customer files into structured data for your product — without manually fixing every new format.
"Customer data cannot leave our infrastructure."
— A security review can stop a deal.
Mildport runs in the cloud — hosted in the EU — or self-hosted.
You decide where your data stays.
“The import worked incorrectly. Why?”
One wrong field can lead to hours of troubleshooting with a customer.
— The ticket nobody wants to close.
Mildport shows how data was mapped and transformed, so your team can review and control the result.
Live demo — try the real Mildport
This is the same import wizard your users will see inside your product — not a video and not a mockup.
Upload a CSV, Excel, or another file and see how Mildport detects data structure, maps fields, and helps prepare the import for your application.
Try these scenarios:
Try real problems SaaS teams run into: extra spaces, duplicates, different phone formats, and unexpected values.
Supported formats:
CSV · TSV · JSON · XML · XLSX
PDFs and scanned documents via OCR.
Your data. Your rules. Your control.
Imagine this import is already embedded in your SaaS.
After the import completes, your application receives exactly the data your process needs — with no hidden steps.
Demo runs entirely in your browser. Uploaded files are not sent to a server.
This demonstrates the same principle Mildport is built on: your data stays where you run it.
Hard imports
Customers don't send perfect files. They send exports from old systems, spreadsheets with different columns, PDF documents, scans, and unexpected formats. Mildport is built for these cases: turning real customer data into structured imports that can be reviewed before reaching your application.
Extract data from documents, scans, and photos. OCR helps turn images into text, tables, and structured data — with review before anything is used.
CSV, Excel, and other formats often arrive with duplicates, missing fields, and unexpected columns. Mildport helps detect structure, map fields, and prepare the import.
Not every customer file is a table. Emails, forms, notes, and documents can become structured records inside your product.
Mildport shows what data was found and how it will be transformed. Users can review, edit, or reject results before anything reaches your application.
Product development
You shouldn't need a sales call to understand what Mildport can do. Here is what works today, what we validate with partners, and what is coming next.
Capabilities available in the current engine.
Scenarios we validate on real customer workflows.
Capabilities we are developing next.
Self-host operations
The most important thing about self-hosted software is not promises. It is predictable operation. Mildport runs where your data lives and gives your team control over deployment, storage, and processing.
Deploy with Docker Compose, or on Kubernetes with Helm.
Files and import state stay inside your infrastructure.
No required data transfer outside your environment. Full control over licensing and delivery.
Health checks, readiness checks, audit trails, and import lifecycle control.
AI features follow your deployment and security rules.
Use cases
Mildport helps when importing becomes part of your product — not a separate tool.
Let customers import their data without building a custom project for every format.
Turn documents, forms, and text into structured records with user review.
Process invoices, receipts, and documents without manual data entry.
How Mildport works
One component adds file import directly into your SaaS. Customers upload data, Mildport helps prepare it, and your product receives structured output.
Add the component to React, Vue, Svelte, or plain HTML. No SDK setup and no need to build an import UI from scratch.
CSV, Excel, PDFs, documents, or multiple files at once. Mildport is designed for real-world customer data — messy, inconsistent, and diverse.
Detects structure, maps fields, highlights issues, and helps users verify results before sending data to your product.
Get clean data via API or webhook in your own schema and processing format.
import { defineImportSuiteElement } from '@capitality-io/mildport-widget';
await defineImportSuiteElement('mildport-import');
<mildport-import
api-base-url="https://imports.your-infra.example"
license-key={SIGNED_TENANT_KEY}
></mildport-import>This is the same component used in the demo above. Embed import into your product and connect it to your data pipeline.
Built like an engine. Delivered as a component.
CSV, TSV, XLSX, ODS, JSON, XML. Extracts tables from PDFs and scans via OCR. Handles large files asynchronously.
Data is processed through deterministic rules first. AI is used only for complex cases and can be disabled or replaced.
Confirmed mappings are stored and reused. The system adapts to how you structure data. Set it once.
Multiple files become a single dataset. Merge by key and work with a unified structure.
Inline editing, field-level validation, error highlighting. Fix issues before data reaches your product.
Rows can reference records in your product. Mildport resolves them using your rules.
Design system support via tokens. The widget feels native to your interface.
Signed webhooks, retries, and delivery logs. Also supports API and browser mode.
Flexible deployment. Full control.
Use Mildport in the cloud or deploy it inside your own infrastructure. Cloud — for a fast start, hosted in the EU. Self-hosted — when data must never leave your environment.
Mildport is built in Germany. In the cloud, your data is stored on EU infrastructure and protected by the GDPR — some of the world's strongest privacy laws. And in self-hosted mode it never touches our cloud at all: files, intermediate results, and final records stay inside your infrastructure.
Deploy Mildport via Docker Compose or Kubernetes. The same engine runs in both cloud and self-hosted setups.
Signed keys are verified locally. No licensing server, no “phone-home” calls. Works in isolated environments.
Every mapping and transformation can be inspected. Decisions are deterministic and audit-ready.
AI is an optional layer for complex cases. You can disable it, replace it with your own model, or run it inside your infrastructure. All decisions are traceable.
# the whole import engine, on your metal
$ cp apps/import-service/.env.example apps/import-service/.env # keys, Mongo, sidecars
$ docker compose up --build
✔ mildport listening on :8090
✔ mongo · pdf-probe · xlsx-decode · ocr
$ curl localhost:8090/health
{"status":"ok","service":"cap-import"}
$ pnpm import:license -- --tenant acme --plan pro
✔ signed key minted — verifies offlineOne engine — both in cloud and in your environment.
For audits and security reviews: data residency & EU hosting → · AI transparency →
Tenant #1: ourselves
In Capitality CRM, data import used to be part of the internal backend. Today it runs entirely on Mildport: the same widget, the same REST API, the same webhook contract. This is a real production migration — with no exceptions or special cases.
commit 98e6161
Date: Sun May 31 2026
chore: cap-auth now consumes the standalone
cap-import service over REST as tenant #1;
the in-process UNP pipeline and ximport
services are removedThis is a real migration commit from the Capitality CRM production repository.
Capitality authenticates with a signed license key against the public REST API and pins the published @capitality-io/mildport-widget from GitHub Packages — semver range, version-pin check in CI. No private endpoints, no special casing. If the public surface is awkward, we feel it first.
Upload, file decoding (XLSX, PDF, OCR), matching, review, apply — all of it runs in the Mildport service now. What remains in the host backend: a tenant-scoped read proxy for the import inbox and a webhook handler that verifies HMAC signatures.
The wizard wears Capitality's design system through --mildport-* tokens, labels flow from the host's own i18n layer, and a recordId deep link resumes a half-finished import straight from the inbox. The embedding surface isn't theoretical — it ships.
Capitality's CI runs Playwright accessibility specs against the embedded wizard. A regression in Mildport turns our product's pipeline red before any customer ever sees it. Every sharp edge you'd hit, we hit first — with production data.
Latest updates
We ship continuously. Here are some recent milestones.
Data is processed with deterministic rules first. AI is used only for complex cases. Each decision includes an explanation and audit trail. The model can be replaced or disabled.
Capitality CRM now uses Mildport as an external import service via REST API. The internal import pipeline has been removed. This migration runs in production with no exceptions.
Upload multiple files and merge them by key (left / inner join). The result is prepared before any data is applied.
Design system support via CSS tokens --mildport-*. Light and dark themes can be enabled without changing logic.
Extract data from scanned PDFs and images. Results go through a confirmation step before being added to the system.
Structured PDFs and documents (e.g. invoices) are parsed and mapped to your schema fields. Preview is available.
Rows can reference existing records in your product. Matching is done using your data rules.
The system uses both headers and value types (email, dates, numbers, etc.) for matching. This improves automatic mapping accuracy.
Deploy with Docker or Kubernetes. Reliable delivery with signed webhooks, retries, and audit logs. Offline licensing and preflight checks included.
React and Angular components are published via GitHub Packages. Versions are pinned via semver and validated in CI.
Product roadmap
We are building Mildport around one principle: making data imports predictable, verifiable, and practical for teams working with real products and real constraints.
The next step after multi-file imports: automatic key suggestions and server-side processing for large XLSX files without browser limitations.
Cloud customers will get separate model, endpoint, and credential settings. Self-hosted deployments keep full control within your infrastructure.
AI will suggest data changes — such as normalizing phone numbers or splitting fields — while every change remains visible and reviewable before applying.
More integration options: framework wrappers and a complete API mode without the embedded interface.
Improving Helm deployment, secrets management, and installation flows for teams requiring full infrastructure control.
Expanding the decision log: mappings, transformations, and manual changes become a reproducible history of every import.
Optimizing validation for large files while keeping the interface responsive and accessible across hundreds of thousands of rows.
Self-service tools for teams: key management, plans, and a simpler onboarding path for developers.
Imports beyond the interface: SFTP, S3, email, and other sources using the same flow — receive, map, validate, and apply.
Enable AI agents to start and control import workflows through APIs inside your environment.
Early access
Commercial access is opening soon. Leave your contact details — we’ll let you know when pilot spots become available and when Mildport is ready to purchase.