Connectors
Publish products from UnoPim, ERPNext, metasfresh, Odoo, Narravero or an Asset Administration Shell to a GS1 Digital Link, a digital product passport and EPCIS events.
Your products already live somewhere: in a PIM, an ERP, a passport platform or an Asset Administration Shell. A connector reads them from there and publishes them to the OpenEPCIS catalog under a GS1 Digital Link, so that a scan of the product resolves to its master data, its passport and its events without anyone typing the data in a second time.
Supported systems
| System | What it publishes | What you set up |
|---|---|---|
| UnoPim | Products, batches | Nothing to install; an optional panel for UnoPim |
| ERPNext | Products, batches, serials, parties, deliveries, receipts, stock entries | An API user, three custom fields and eleven webhooks, created by one setup script |
| metasfresh | Products, parties | A read-only API token |
| Odoo | Products, contacts, lots, serials, movements, packing, production, sales | Seven addons, installed in Odoo |
| Narravero | Products | An API token, a view, the context ID and the payload path of the GTIN |
| AAS: BaSyx, EDC | Products, batches, serials, documents | Configured with you on request |
Odoo is the one system where the connector lives inside the application itself: seven addons that publish on a schedule, report what happens in the warehouse, in production and at the till, and take care of single sign-on logout. They are open source under LGPL-3 in openepcis-odoo. For all other systems we run the connector as a service that polls your system, and you only hand over access. UnoPim users can add a small panel to their PIM that shows the publishing status next to each product; it is published under MIT as openepcis-unopim. For ERPNext, the setup script adds a similar panel to the Item form.
If you are new to this, how a connector works explains the mechanics, and a warehouse day in Odoo shows the whole thing screen by screen.
What a published product looks like
Every product gets a master data document at /01/<gtin>, and a batch or a single item gets its own at /10/<lot> or /21/<serial>. The document is JSON-LD over the GS1 Web Vocabulary, which means it expands to RDF and can be validated against SHACL shapes, and you can extend it with terms from the battery, textile, packaging or EUDR vocabularies.
From that document the platform derives an EN 18223:2026 passport whenever someone reads it, in JSON (compressed, full or gs1), XML, Turtle or N-Triples. A passport needs a Digital Link and an economic operator, so a product that arrives without a manufacturer ends up in the catalog and gets its passport once the manufacturer is filled in.
Where the source system knows about individual items, the connector also writes EPCIS events: a batch as a class, a serial number as a single unit, a pallet under its SSCC. The event ID is the CBV hash of the event, so sending the same event twice has no effect. All of this hangs together in a linkset, which is what a scanner reads when it resolves the Digital Link.
Who sees which fields
Each field belongs to one of three tiers. Public fields are visible to everyone, AuthorizedOnly fields to any member of the owning tenant, and Restricted fields only to members with the restricted role. Fields that nobody marked count as AuthorizedOnly. Access to whole documents is enforced today; filtering inside a document is controlled by dlr.field-access.enforce, which is off by default.
Standards note: EN 18222 and EN 18223 are two of six parts CEN/CENELEC JTC 24 published in 2026 under request M/604. EN 18239 and EN 18246, which cover access rights and data authentication, followed in August 2026. The series is not yet cited in the EU Official Journal, so implementing it gives no presumption of conformity with the ESPR.
Before you switch it on
Each connection has its own mapping, a YAML file that says which source field becomes which GS1 term and which side owns it. Connections start in dry-run and only log what they would write, which is worth reading carefully, because once you switch it off the catalog writes to real registrars, GS1 among them. Publishing the same record twice changes nothing, and webhooks from your system only tell the connector to read a record again; their content is never taken over as data.
A few things behave differently from what you might expect. Deleting a record in your system leaves the catalog entry in place. Corrections made on the platform side reach your system on the next poll. Odoo and ERPNext report movements, receipts and production as EPCIS events, while the other connectors report commissioning. UnoPim, Odoo and ERPNext show the publishing status inside your own application.