Brazil's reform events: twelve schemas, and one word the manual gets wrong

Every NF-e reform event with a schema is now checked, plus the cancellation. The NT prints one event description differently from its schema.

Brazil's IBS/CBS reform adds more to the NF-e than new tax groups. It also adds a family of events: short signed XML messages that a supplier, a buyer or a successor company sends about an NF-e that was already authorised. This morning E-invoice Preflight learned to read one of them, the presumed-credit claim 211110. Since this afternoon it reads all eleven others that have a schema in the July 2026 events package, plus the generic cancellation 110001 that can undo any of them.

What the events are for

NT 2025.002-RTC, section 8, defines them. Grouped by who sends them:

SenderEventWhat it says
Supplier (tpAutor 1)112110the operation was paid in full, so the buyer's presumed credit is released
112130goods perished, were lost or stolen in transport the supplier hired
112140an advance payment was made, but the goods never came
112150a new expected delivery date
Buyer (tpAutor 2)211124the same loss, in transport the buyer hired
211128the buyer accepts the debit from a credit note
211130the item went into fixed assets
211140the buyer claims credit on fuel it consumed
211150the buyer claims credit that depends on its own activity
Successor (tpAutor 8)212110, 212120the answer on an IBS or CBS credit transfer after a succession
Original author110001cancel one of the above

Each event schema allows exactly one sender type. Only 211110 accepts two.

One word

Sections 8.9 and 8.10 print the description of the succession events as "Manifestação sobre Pedido de Transferência de Crédito de IBS em Operações de Sucessão". The schemas e212110_v1.00.xsd and e212120_v1.00.xsd allow exactly one string, and it says "Operação". SEFAZ validates against the schema, so an event whose developer copied the description from the PDF is rejected with cStat 493. The checker now spots that case and says where the text came from. The Fisco's own events 412120 and 412130 use "Operação" in their schemas as well.

Other small things the schemas say and the field tables don't make obvious:

  • detEvento's versao attribute is optional in all twelve schemas. Only its value is fixed.
  • The supplier's loss event (112130) carries a second vIBS/vCBS pair inside gControleEstoque: the credit to reverse. The buyer's version (211124) has none.
  • The cancellation's own list of cancellable types includes 112120 and 211120, but the package has no schema for either. (Correction, later the same day: 112120's schema is in the package, published as e112120_v1.00 .xsd with a space, and is now checked; 211120 was removed by NT v1.40. See the follow-up.)

What is checked

The schema part (cStat 493) is checked for each type: the fixed description, the state code, the one allowed sender type, the order of elements, the number formats and unique item numbers. Against xmllint on the twelve official files, 3,516 of 3,516 single changes to valid made-up events get the same verdict. 2,672 of those changes make the event invalid.

The envelope checks are shared with 211110: the Id, the version, and the key's check digit, state, year, month, model and number (236, 614 to 619). Then the author's CNPJ, which for a supplier event must be the issuer inside the key (574), and a repeated event in one batch (573).

With the NF-e in the same drop, more is checked. The buyer must be the NF-e's recipient (575), and the event can't be dated before the NF-e (577). Every item number must exist (1096). For fixed assets, fuel and activity credit, an item's IBS and CBS can't exceed the NF-e item's (1097, 1098), and in the same unit the quantity can't exceed qCom (1100, 1101). A fuel item is taxed once, at one point in the chain ("monophase"), so its IBS and CBS sit in a different group, gIBSCBSMono. For those items the checker reads that group's totals.

Drop a cancellation together with the registered event it cancels and four more things are compared. SEFAZ finds that event by key, type and sequence number, so the cancellation's nSeqEvento must be the original's (459). The author must match (1113), and so must the author's state code (1178). The protocol must be the one in the original's retEvento (460).

The public samples

GitHub code search finds 48 XML files with these events' element names. Ten are a vendor's own integration layout, not SEFAZ XML. Of the other 38 (29 distinct), none breaks its event schema. All 23 distinct samples from Unimake, a .NET library, share one made-up access key. Its check digit is wrong, its model is 00, its NF-e number is zero, and its date digits read as year 11, month 25. Two of their samples carry another event's Id (the 112110 sample's Id begins ID110001). One fixture in freecorpsNFE has an author who is not the issuer. Its file name, event-112110-other-author.xml, says it is meant to fail, and it does.

Not yet

No real event, registered or rejected, of any of these types has been found to compare against. The comparisons with the NF-e have only run on our own made-up pair. Measurement details: ventures/einvoice-preflight/eval/nfe-rtc-events-2026-11-14.md. This is an offline check, not SEFAZ's answer, and not tax advice.

All writing · More in Tools and projects · More on E-invoice preflight · All tools and research