Reports & oppositions
This guide covers the registrability- and opposition-focused reads of Tarno's contract and the
one-call reports (/reports/*). The four reads use the brands:read scope; the reports carry
their own permission (see Scopes). Everything has an MCP twin.
The complete, always-current per-field schema lives in the interactive OpenAPI reference at
https://api.tarno.cl/docs(Scalar). This guide explains how to request these reads and links to that reference — it does not restate the schema. The reference is authoritative for the field list.
Throughout this guide the base URL is https://api.tarno.cl and every call carries the X-API-Key
header.
What this guide covers
Four reads, all without a version prefix and with the brands:read scope:
- Registrability report — no longer a route: it is the trademark detail with
expand=parties,annotations,encumbrances,classes,norms. One object with one null semantics. - Opposition timeline — no longer a route: it is the
family=oppositionfilter over a mark's annotations. - Open oppositions — no longer a route: it is search with
oppositionWindow: "open". - Renewal calendar —
GET /trademarks/renewal-calendar≡get_renewal_calendar: the renewal dates (art. 24) of a batch or a filter in a range of up to 366 days (details).
And the one-call reports (the most recent ones have their section below), with their own permission (reports:read or the report's own):
- Daily bulletin —
GET /reports/daily-bulletin≡get_daily_bulletin: what the Estado Diario published on a date, by section and with parties. - Clearance —
GET /reports/clearance≡get_clearance_report: what resembles a name in one or more classes, as signals (never "registrable"). - Portfolio of a party —
GET /reports/portfolio≡get_portfolio_report: a party's marks (by itsid, never by its name) with their counts, upcoming milestones, opposition windows, oppositions received, representatives and relations. - Brief to file an opposition —
GET /reports/opposition-brief≡get_opposition_brief: the target mark, whether it is in its window today, its similarity to your mark and its holders' opposition history. - Brief to respond to an opposition —
GET /reports/opposition-response-brief≡get_opposition_response_brief: the opposition acts on your mark, who opposes, with which norms and with which history. - Facts of a filing —
GET /reports/filing-context≡get_filing_context: the facts of a power of attorney, an opposition, an answer to an opposition, an answer to observations or a withdrawal, each with its source, no drafting (details).
⛔ They all give signals, never verdicts: none says whether a mark is registrable, whether an opposition will succeed or whether there is a risk — the legal decision is the reader's.
These reads are read-only and honest by design: they never fabricate dates or deadlines the corpus cannot sustain.
Registrability report
GET /trademarks/{applicationNumber}?expand=parties,annotations,encumbrances,classes,norms — scope
brands:read (plus legal:read when expand includes norms). MCP: get_trademark with expand.
The report is the trademark detail with its sections expanded: parties, annotations,
encumbrances, classes and norms. One object with one null semantics — the defect of two routes
(detail and report) with the same shape and two semantics is closed.
⚠ null is NOT [], and the difference is the whole mechanism:
| value | means |
|---|---|
[] |
I asked, and this mark has none |
null |
I didn't ask (not in expand=) |
Encumbrances (encumbrances)
encumbrances[] carries one entry per charge over the mark —pledge, precautionary measure, attachment
or prohibition— with its registration paired to the release that lifted it. Active ones first.
state |
meaning |
|---|---|
active |
registered and not closed by any registered release |
released |
a registered release closed it (recordedOn: null when its registration predates the corpus) |
filed |
filed, no resolution closing it yet (today, 2024–2026 filings in process) |
rejected |
INAPI rejected it (537); never in force |
abandoned |
abandoned without registration (non-payment, unmet observations or withdrawal); never in force |
A charge is in force only once INAPI registers it: the A0xI annotation is the filing, and what makes
it rule is the registration resolution (1104, 638, or its publication in the Estado Diario). A release
(A0xA) lifts only once registered (639). filed, rejected and abandoned are not encumbrances in
force and should not trigger alerts; they carry filedOn (the filing date) and recordedOn: null.
Charges from 2007–2011 (registrationSection W340, W360 or W370) always carry filedOn, the filing
date; recordedOn is their registration date. On a modern charge in force or released, filedOn is
null and recordedOn is the date of its A0xI. ⚠ With releaseObservable: false an active charge
may have been lifted: the mark carries, after its registration, a release the corpus cannot attribute to
one charge (INAPI's file names the annotation number; the corpus does not keep it). Those releases are
listed apart as released with recordedOn: null.
recordedOnBasis says which date recordedOn is: "filing" on a modern charge (the filing date of
its A0xI; the registration is a later resolution that is not published as a date), "recordal" on a
2007–2011 charge (its registration), null when recordedOn is null. Do not print "recorded on…" over a
"filing" date.
⚠ With releaseAttributionCertain: false, which charge a release lifted is an attribution, not a
fact: the release arrived while two or more charges of its kind were in force and was paired with the
oldest. Every charge in force at that moment carries it; the one listed released may still be in force
and an active one may be the one lifted. How many releases there were is certain; which charge each one
lifted is not.
releaseStatusCertain answers a different question: whether the state (active / released) is
certain even though the pairing is not. It is true when the attribution is already certain, when every charge of the kind on the mark is released
(each one is, for certain; only its releasedOn is an attribution) or when the release could only have
lifted one charge, because the other was filed that same day. It is false whenever releaseObservable
is false. To say "still in force" or "released", use releaseStatusCertain; to say "released on…", also
releaseAttributionCertain. Where the history carries INAPI's annotation
number, two identical charges of one day are two entries, and when INAPI states which charge a release
lifts the pairing is certain. Limit: on a mark whose history carries no numbers yet, two identical
charges of one kind filed on the same day can show as one, and then the state can be wrong.
Where INAPI's resolutions carry the annotation number, each filing is resolved by its own number: a
release filed twice on one day whose duplicate was withdrawn or rejected counts once, and a charge
rejected or abandoned by its number is not published as in force even if another of the same day was
registered. If a release's registration resolution (639) names a charge that is not in force on the
mark, that release never takes a charge another 639 names, and whatever it closes by age, and every other charge it could have taken, carries
releaseAttributionCertain: false and, unless every charge of the kind is released,
releaseStatusCertain: false.
Each entry also carries annotationNumber, INAPI's annotation number of the charge (the one cited),
and releaseAnnotationNumber, that of the release that lifted it; null when unknown.
The norms cited in the file (norms)
norms[] carries one entry per (kind, norm, article, letter, subsection) cited in the trademark's
annotations: normKind, legalNormId (null when the act cites an article without naming a law — still
included), citedNorm, name, article, letter, subsection, annotationCount, firstCitedOn,
lastCitedOn and up to 20 annotationIds (the most recent). Most cited first, at most 100
(normsTruncated). With dedupe=true (the default) it counts facts, and the ids are those of
annotations[]. It does not carry the law's text: to read it, GET /annotations/{id}?expand=citations
with one of the annotationIds.
⚠ norms requires the legal:read scope in addition to brands:read, like expand=citations: without
it the whole request answers 403 forbidden naming legal:read, and no quota is consumed. A key without
legal:read asks for expand=parties,annotations,encumbrances,classes.
curl -sS "https://api.tarno.cl/trademarks/123456?expand=parties,annotations,encumbrances,classes,norms" \
-H "X-API-Key: $TARNO_API_KEY"
const res = await fetch(
"https://api.tarno.cl/trademarks/123456?expand=parties,annotations,encumbrances,classes,norms",
{ headers: { "X-API-Key": process.env.TARNO_API_KEY! } },
);
if (!res.ok) {
const problem = await res.json(); // application/problem+json (RFC 9457)
throw new Error(`${problem.status} ${problem.code}: ${problem.detail} (requestId=${problem.requestId})`);
}
const trademark = await res.json(); // trademark object with the requested sections
console.log(trademark.status, trademark.statusInapi, trademark.classes); // status = derived (D11)
The act of each annotation (annotations[]): besides section, every annotation carries act —the
INAPI code of the concrete act it names (e.g. "Fondo-Res. Favorable Escrito" under M11 is
664B)— with actName/actLabel, and actStatus, which always travels and says what a null
act means: not_applicable (the section already is the act), unidentified (it names an act that
could not be identified: "we do not know", never a plausible code) or identified.
For the full list of trademark fields (and each expanded section), see the reference
https://api.tarno.cl/docs — not duplicated here. Freshness travels in headers (Iris-Data-As-Of,
Iris-Coverage-Through) and ETag + If-None-Match → 304.
A mark's opposition timeline
GET /trademarks/{applicationNumber}/annotations?family=opposition — scope brands:read. MCP:
list_trademark_annotations with family: "opposition".
The opposition timeline is the family=opposition filter over a mark's annotations: a chronological
collection of opposition acts, with the common envelope { results[], limit, next, hasMore, total, meta }.
curl -sS "https://api.tarno.cl/trademarks/123456/annotations?family=opposition" \
-H "X-API-Key: $TARNO_API_KEY"
The opposition window (publication + 30 business days) lives in the trademark object
(oppositionWindow), not in this collection. null means the mark has no publication date, so there is no
window to compute — never fabricated without its origin date.
The mark's dated milestones travel in milestones[] (on the trademark object and on every mark of a
collection); the list of kinds, what each publishes and when it is an estimate, is in
Milestones. The close of this window travels there as opposition_close, with its
legal basis (Ley 19.039 arts. 5 and 11, Ley 19.880 art. 25) —derived, or estimated and flagged when
it cannot be affirmed—, and oppositionWindow carries the same legalBasis.
Open oppositions in a class
POST /trademarks/search with oppositionWindow: "open" — scope brands:read. MCP: search_trademarks.
Lists marks with an open opposition window, ranked by similarity to a portfolio. The use case: watch
for registration threats in your class. The portfolio to rank against is q (free text) or
applicationNumber (ONE existing mark of yours), and the class is required:
curl -sS -X POST "https://api.tarno.cl/trademarks/search" \
-H "X-API-Key: $TARNO_API_KEY" \
-H "Content-Type: application/json" \
-d '{"oppositionWindow":"open","niceClass":[43],"q":"cafe example"}'
The window's rules.
niceClassis required (one or more classes) and you must send exactly one ofqorapplicationNumber(a single number: it is the mark to rank against, not a batch). ⛔ The window does not combine withsimilarTo/similarToTrademark— it already ranks by similarity to the portfolio —: that combination is a400 validation_error. For "marks similar to X" without the window, usesimilarTo(free text) orsimilarToTrademark(an existing mark), never both.
oppositionWindowaccepts only"open":closedandanydo not exist and are a400 validation_error. A portfolioapplicationNumberthat is not in the corpus (or is not even number-shaped) is a404 not_found, never an empty200: "nothing similar in the window" is only said about a mark that exists.
Which mark is "in the window". One that (1) has a publication date (publishedOn, the Diario Oficial
publication), (2) whose 30th business day after publication is today or later — Chilean days,
counted with the holiday calendar — and (3) whose status is not terminal (rechazada,
denegada, desistida, abandonada, anulada, caducado, vencida; the liveOnly list) and (4) whose status
is not registrada or concedida either — a registered or granted mark no longer admits opposition
(liveOnly, by contrast, keeps it: it is alive). It uses the
same term the trademark object publishes in oppositionWindow, with the same arithmetic: not a
separate deadline. But oppositionWindow publishes the term without looking at the status: a
registrada/concedida mark can carry expired: false and not be in the window — do not infer
membership from !expired. In practice these are en_tramite marks — INAPI never emits the publicada status. If
the calendar does not cover today − 90 to today + 60 days, the answer is 422 out_of_coverage, never a list computed
with missing holidays. The same rule drives the "in window" flag of the clearance and portfolio reports
(inOppositionWindow).
The other filters are applied (no longer silently ignored): status, signType,
holder/representative/taxId (with relation), the four date pairs (filedFrom/To,
registeredFrom/To, expiresFrom/To, publishedFrom/To), registrationNumber, liveOnly, and
multiple classes (niceClass as a list). The opposition window also accepts multiple classes. They
are applied inside each similarity channel, not as a post-filter (except in the lexical channels with
liveOnly, status=caducado, expiry, milestones or renewalStatus, where they apply after scoring: same
result and same order).
sort only accepts relevance: results are ranked by similarity, and sorting by date or
denomination would destroy that order. denomination and facets do not combine with similarity or the
opposition window. Any of these combinations is a 400 validation_error.
And the same errors as the collection: a taxId that is not shaped like a RUT is 400 invalid_rut,
and a range with From > To (filedFrom/To, expiresFrom/To, registeredFrom/To, publishedFrom/To,
milestoneFrom/To) is 400 invalid_range — with the same message. Never a 200 with an empty list.
liveOnly=true excludes marks in one of the seven terminal statuses (rechazada, denegada, desistida, abandonada, anulada, caducado, vencida); a mark
with no known status is kept.
The response is the common envelope { results[], limit, next, hasMore, total, meta }, paginated by
keyset.
Renewal calendar
GET /trademarks/renewal-calendar?from=…&to=… — scope brands:read. MCP: get_renewal_calendar.
The renewal dates of a set of marks in one call: one row per (mark, milestone) whose date falls in
the range, ordered by date and application number. The milestones are the two of the renewal window of
art. 24 of Ley 19.039 —renewal_window_open (expiry − 6 months) and grace_end (expiry + 6 months, moved to the next business day when it is a non-business day)— and
every row is the milestone as the mark publishes it in its milestones[] (same date, anchor, rule and
legalBasis), plus its applicationNumber.
| parameter | what it is |
|---|---|
from / to |
required (YYYY-MM-DD, inclusive); at most 366 days counting both ends. More is 400 validation_error; from > to is 400 invalid_range |
| trademark filters | those of GET /trademarks: the batch (applicationNumber, up to 100), renewalStatus, holder, niceClass, status, dates… denomination narrows by substring. q, similarity, the opposition window, facets and the norm filters are 400 |
limit / cursor |
the usual pagination (default 20, max 100); total is exact. A cursor is only valid with the same filter (range included; limit may change): with another, 400 validation_error on cursor |
A mark of the pre-2022-05-09 regime, filed via Madrid, without an expiry date or neither registrada nor
esperando_renovacion does not publish those milestones and contributes no rows; nor does one in
renewalStatus: renewed_expiry_pending (the renewal is already granted and the published expiry is not
yet updated, before or after that expiry). The calendar
publishes dates; who to alert and when is the consumer's decision.
# "The renewal dates of my list this year" — the batch of up to 100, in one call.
curl -sS "https://api.tarno.cl/trademarks/renewal-calendar?applicationNumber=1161630,1197627,1192677&from=2026-01-01&to=2026-12-31" \
-H "X-API-Key: $TARNO_API_KEY"
# → results = [{ "applicationNumber": "1161630", "kind": "renewal_window_open", "date": "2026-01-18",
# "basis": "derived", "anchor": { "kind": "expiry", "date": "2026-07-18" },
# "rule": "6_months_before_expiry", "legalBasis": [ … art. 24 … ] }, …]
{ "name": "get_renewal_calendar", "arguments": { "applicationNumber": ["1161630"], "from": "2026-01-01", "to": "2026-12-31" } }
The portfolio as an object: POST /trademarks/renewal-calendar. The same answer by the JSON body, plus
the portfolio object (party, refine, includeApplicationNumber, excludeApplicationNumber), the same
one as POST /trademarks/search and POST /trademarks/application-numbers: set =
(((parties ∩ refine) ∪ included) − excluded) ∩ top-level filters. refine narrows only the parties'
marks, so a mark added by hand is in the calendar even when it fails refine, and one excluded by hand
never is. Up to 10,000 numbers per list. With portfolio, a top-level party, includeApplicationNumber
or excludeApplicationNumber is a 400 validation_error. The cursor is bound to the whole filter,
portfolio included. get_renewal_calendar takes the same portfolio; the GET stays flat.
curl -sS -X POST "https://api.tarno.cl/trademarks/renewal-calendar" \
-H "X-API-Key: $TARNO_API_KEY" -H "Content-Type: application/json" \
-d '{ "from": "2026-01-01", "to": "2026-12-31",
"portfolio": { "party": ["123:holder"], "refine": { "niceClass": [9] },
"includeApplicationNumber": ["1161630"], "excludeApplicationNumber": ["1197627"] } }'
Daily bulletin
GET /reports/daily-bulletin?date=YYYY-MM-DD — permission reports:read or reports:daily-bulletin:read.
MCP: get_daily_bulletin.
What the Estado Diario published on a date, in one call: every act of the day in the published
annotation shape (parties included), in the common paginated envelope (limit ≤ 100, cursor), with an
exact total. In meta:
| field | what it is |
|---|---|
meta.date |
the requested date |
meta.sections[] |
the whole day's index: { code, label, count } per section, label in es-CL/en/pt-BR — it does not change when you filter by section |
meta.bulletins[] |
the bulletins ingested that day (tipo2, altas) with the acts the ingestion recorded |
meta.coverageThrough |
⭐ the latest ingested Estado Diario bulletin: the same value as coverage.through of the 422 out_of_coverage (not the corpus filing-date frontier used by every other route) |
section (a list, ?section=M5,OPO08 or repeated) narrows the rows. Rows are not collapsed by
event: the bulletin is what was published; versionsOfFact still says when a fact has other versions.
Every row also carries act and actStatus, as in the report: under the subsection heading the
bulletin prints ("Fondo-Res. Favorable Escrito" under M11) act carries the INAPI code of that act
(664B); actStatus says whether it is identified, whether the bulletin names an act that could not
be identified (unidentified), or whether the section already is the act (not_applicable).
⛔ A day with no ingested bulletin is 422 out_of_coverage, with coverage: { from, through } in the
error body — never an empty list: "nothing happened" and "we did not read it" are different answers.
Weekends and holidays have no bulletin either; GET /system/coverage tells a holiday from a gap.
curl -sS "https://api.tarno.cl/reports/daily-bulletin?date=2026-09-22§ion=M5&limit=50" \
-H "X-API-Key: $TARNO_API_KEY"
{ "name": "get_daily_bulletin", "arguments": { "date": "2026-09-22", "section": ["M5"], "limit": 50 } }
Clearance (signals, not verdicts)
GET /reports/clearance?text=…&niceClass=… — permission reports:read or reports:clearance:read. MCP:
get_clearance_report.
What resembles a name in one or more classes (niceClass, 1 to 10, required), in one call, as a set of
signals with their source. ⛔ It does not say whether the name is registrable, available, risky or
infringing: that is a legal opinion, and it is the reader's. No report publishes such a field (a contract
test forbids it).
| field | what it is |
|---|---|
similar[] |
the confusable marks by channel of the similarity engine: phonetic, fuzzy (lexical; omitted when its channel failed on this report), semantic (omitted when it was not evaluated or contributed nothing — never empty) and distinctive (the distinctive element, without the generic terms; only when the channel ran), up to 20 per channel, each with its score/signals and inOppositionWindow |
genericTerms, distinctiveCore, decompounded |
only when the distinctive channel ran: which terms were treated as generic and why, what was searched and which joined words were split |
dominantHolders[] |
the 10 holders with the most marks in those classes ({ key: partyId, label, count }, any status) |
citedNorms[] |
the norms/articles cited in the confusable marks' annotations, with trademarkCount, annotationCount and up to 5 evidence numbers (closest first) |
dataAsOf, coverageThrough |
freshness, in the body (and in headers) |
inOppositionWindow is membership (the mark is inside its 30 business days of opposition today);
the closing date travels in the mark's milestones[] as opposition_close, with its article. A cited norm is a signal of
what has been invoked around similar marks, not a ground applied to your name.
The distinctive group («Similarity in the distinctive element»). When the text carries generic terms
in the requested classes («ALTODESEMPEÑO CONSULTORES» in 35), the report adds a group with what resembles
what is left: in that case it finds «MAS DESEMPEÑO», which the other three channels missed. How a term is
judged generic, and how much the channel weighs, is in Search. With nothing to
discount, or when what is left has fewer than 3 letters, there is no group and the three keys are absent:
the report is what it was before. Latency: with generic terms the report takes
about 0.7 s longer (measured on 2026-09-28). ⛔ A term being generic is not a verdict.
curl -sS "https://api.tarno.cl/reports/clearance?text=cafe%20andino&niceClass=43,30" \
-H "X-API-Key: $TARNO_API_KEY"
{ "name": "get_clearance_report", "arguments": { "text": "cafe andino", "niceClass": [43, 30] } }
Conflict landscape
GET /reports/conflict-landscape?text=…&niceClass=… — permission reports:read or
reports:conflict-landscape:read. MCP: get_conflict_landscape.
"Who represents, and who holds, the trademarks similar to mine?", in one call. It starts from the
same engine and channels as clearance (phonetic, lexical, semantic and distinctive
element): the 60 trademarks most similar to the text in the classes (niceClass, 1 to 10,
required). Of those it keeps the live ones: terminal statuses are counted and left out. Then it groups
them by the parties with a current relation.
| field | what it is |
|---|---|
representatives[], holders[] |
up to 20 parties per side (those with the most similar trademarks first) |
└ party |
{ id, name } — the id of /parties |
└ similarTrademarks |
how many of the live similar trademarks the party has, with a current relation |
└ evidence[] |
up to 5 of those trademarks (closest first), with applicationNumber, denomination and matchedBy; evidenceTruncated if there are more |
└ totalTrademarksInClass[] |
for each requested class, all the party's trademarks on that side (any status): the size to read similarTrademarks against |
partiesTotal |
how many parties each list had before the cut at 20 |
method |
the threshold (similarLimit = 60), how many the engine returned, how many are live and how many terminal ones were left out, the terminal statuses and relation: current, with the description in three languages |
genericTerms, distinctiveCore, decompounded |
the engine's discount, only when the distinctive channel ran (as in clearance) |
dataAsOf, coverageThrough |
freshness |
⛔ They are signals from the corpus. They say nothing about a party, a conflict or the registrability
of the name. A representative with many similar trademarks in a class may simply be big: that is why
totalTrademarksInClass travels with it.
Its own permission: it is built on clearance, which is a report that is sold, so it has its own
permission. Neither reports:clearance:read nor a key without restriction gets it. Without it,
403 forbidden naming reports:conflict-landscape:read, with no quota consumed.
curl -sS "https://api.tarno.cl/reports/conflict-landscape?text=altodesempe%C3%B1o%20consultores&niceClass=35" \
-H "X-API-Key: $TARNO_API_KEY"
{ "name": "get_conflict_landscape", "arguments": { "text": "ALTODESEMPEÑO CONSULTORES", "niceClass": [35] } }
Orientative index (signals, not a verdict)
GET /reports/registration-signals?text=…&niceClass=… — permission reports:read or
reports:registration-signals:read. MCP: get_registration_signals.
"Which conflict signals does this name have in these classes, and where does each one come from?", in
one call. It is not a "risk" with a level: it is a breakdown of signals, each with its value, its
n, its evidence and its source, and a signalIndex from 0 to 100 whose formula travels in the
answer. niceClass is required, 1 to 3 classes (the report runs one clearance and the statistics
of each class; with 4 ⇒ 400).
For each class (perClass[]), six signals:
| signal | what it measures | where it comes from | how it is normalized |
|---|---|---|---|
| S1 | the highest similarity with a live trademark of the class | clearance (one call, all the classes): the score of the live similar trademarks |
max(score); 0 when there is none |
| S2 | how many live trademarks are very similar | clearance: those with score ≥ 0.8 |
min(1, log10(1+n)/log10(21)): reaches 1 at 20 |
| S3 | the share refused among the decided applications of the class | get_class_stats, default period |
refusedShare = refused / decided |
| S4 | the share with a substantive objection in the class | get_class_stats |
actsShare.substantiveObjection.share |
| S5 | the share of the live similar trademarks with an opposition act | the oposicion-family annotations of those trademarks |
with an act / live similar, with its coverageNote |
| S6 | the live similar trademarks in an open opposition window today | clearance (inOppositionWindow) |
none: a count, outside the index |
Each signal carries id, name (three languages), value, n, inIndex, weight, nullReason,
source, normalization and evidence (up to 5 trademarks in similarity order, or the share with its
period and dataAsOf). Live = a published status outside the terminal ones (method.terminalStatuses),
the same rule as the conflict landscape.
The formula (field formula, literal, and weights):
signalIndex = round(100 × (0.35·S1 + 0.20·S2 + 0.20·S3 + 0.15·S4 + 0.10·S5))
- A null signal (
value: null, with itsnullReason:class_stats_pending,no_decided_in_period,no_population_in_period,no_live_similar) is left out and its weight is shared among the others in proportion to their weights.weightsAppliedgives the applied weights (rounded to 4 decimals), and the index is reproduced exactly from them:round(100 × Σ weightsApplied·value). - The weights are provisional (
method.weightsProvisional: true): approved by the Tarno team for review, not the result of a calibration. They are reviewed with use. indexLabel: "Orientative index: not a verdict", in three languages. Quote it with the index.
⛔ What it is NOT. It does not say whether a name can be registered, nor whether it infringes, nor a level (low, medium, high), nor a recommendation to file or not. These are signals from the corpus with their source; the legal decision belongs to the reader. Show the signals, not only the number.
Before the first computation of the class statistics (right after a deployment), S3 and S4 come back
null with class_stats_pending and their weight is shared: the index is computed from S1, S2 and S5, and
says so.
Own permission: it is built on clearance, which is a report that is sold. Neither
reports:clearance:read nor a key without restriction gets it. Without it, 403 forbidden naming
reports:registration-signals:read, with no quota consumed. Latency: that of clearance (1–4 s).
curl -sS "https://api.tarno.cl/reports/registration-signals?text=altodesempe%C3%B1o%20consultores&niceClass=35" \
-H "X-API-Key: $TARNO_API_KEY"
{ "name": "get_registration_signals", "arguments": { "text": "ALTODESEMPEÑO CONSULTORES", "niceClass": [35] } }
Portfolio of a party
GET /reports/portfolio?partyId=… — permission reports:read or reports:portfolio:read. MCP:
get_portfolio_report.
The portfolio of one party in one call. ⛔ It is built from the party's identity (partyId: the
id of GET /parties/{id} / get_party, or an annotation's partyId), never from its name: holder
is a substring, and "ACME"'s portfolio by name would bring in "ACME CHILE"'s marks.
| parameter | what it is |
|---|---|
partyId |
required — the party's id |
relation |
current (holder today, default), former (was) or all; selects the portfolio's marks; the relation sections apply it as list_party_relations does (on both bridges of each edge: with former, a representative replaced on a mark the party still holds counts) |
horizonDays |
how many days ahead to look for milestones: default 90, maximum 366 |
| field | what it is |
|---|---|
party |
{ id, name, taxId, kind } |
trademarkCount, byStatus[], byClass[] |
how many marks, and the count by published status and by class ({ key, label, count }) |
byRenewalStatus |
⭐ { groups, undatedCount, notApplicableCount }: the marks by renewalStatus ({ key, label, count }, only the statuses present), those without an expiry date and those with a date the rule does not apply to (Madrid, or neither registrada nor esperando_renovacion). The three add up to trademarkCount; it is the same aggregate as groupBy=renewalStatus |
upcomingMilestones |
from/to (today and today + horizonDays) and the marks with a milestone in that range, soonest first, each with its full milestones[]. ⚠ The portfolio lists marks by their expiry (expiry): a mark whose only milestone in the horizon is its opposition_close is not listed here (it shows in inOppositionWindow); up to 50 + hasMore |
inOppositionWindow |
the marks today inside their opposition window — membership; the close is in each mark's milestones[] as opposition_close; up to 50 + hasMore |
oppositionsReceived |
who opposed (the opposed_by relation edges, up to 20), unresolvedOpponentMentions (mentions that match no single party: they exist, they are not named) and coverageFrom (oppositions exist in the corpus only since that date) |
representatives |
its representatives (represented_by edges, up to 20) |
relations[] |
one row per relation type with partyCount and trademarkCount |
dataAsOf, coverageThrough |
freshness, in the body (and in headers) |
⭐ Every mark carries renewalStatus (not_due | window_open | grace | lapsed | renewed_expiry_pending, or null when
the rule does not apply: no expiry date, neither registrada nor esperando_renovacion, or filed via Madrid), evaluated on the date of the
response (Chile time), and its milestones[] carry the renewal milestones renewal_window_open and
grace_end (Ley 19.039, art. 24; none in renewed_expiry_pending) — the rule is in Milestones; byRenewalStatus
counts them with the same rule and the same "today". The renewal dates of the whole portfolio in a range
come from the calendar with holder or the batch. No verdicts and no opposition closing dates outside
milestones. An unknown party is 404 not_found.
curl -sS "https://api.tarno.cl/reports/portfolio?partyId=920&horizonDays=180" \
-H "X-API-Key: $TARNO_API_KEY"
{ "name": "get_portfolio_report", "arguments": { "partyId": "920", "horizonDays": 180 } }
Brief to file an opposition
GET /reports/opposition-brief?applicationNumber=… — permission reports:read or
reports:opposition-brief:read. MCP: get_opposition_brief.
What you need to prepare an opposition against a mark, in one call, as signals with their evidence.
| parameter | what it is |
|---|---|
applicationNumber |
required — the target mark |
ownApplicationNumber |
optional — your mark; without it there is no similarity signal |
limit |
cap of each list (holders, edges): default 10, max 20 |
| field | what it is |
|---|---|
trademark |
the mark as the collection publishes it (classes, status, milestones[]), plus publishedOn verbatim and currentHolders/currentRepresentatives by id |
inOppositionWindow |
true if it is inside its opposition window today — membership; the close is in trademark.milestones as opposition_close |
similarity |
only with ownApplicationNumber: where the target ranks among the 100 marks most similar to yours in its classes (rank, score, matchedBy, signals), or found: false. Without your mark the key is absent (never filled) |
holders |
each current holder of the target with oppositionsFiled (whom it opposed: opposed edges) and oppositionsReceived (who opposed it: opposed_by edges, plus unresolvedOpponentMentions), each with partyCount, trademarkCount and the most frequent edges |
citedNorms |
the norms cited in the opposition acts received by those holders' current marks, with counts and up to 5 evidence application numbers (the target first) |
coverageFrom |
oppositions exist in the corpus only since that date |
dataAsOf, coverageThrough |
freshness, in the body (and in headers) |
⛔ No opposition closing date outside milestones (it travels in milestones as opposition_close, with its article), no verdict,
no renewalStatus (a brief does not publish a state that changes every day; the renewal-window dates do
travel in milestones). Unknown target or own mark: 404 not_found; the same mark in both parameters: 400.
curl -sS "https://api.tarno.cl/reports/opposition-brief?applicationNumber=1712345&ownApplicationNumber=1698765" \
-H "X-API-Key: $TARNO_API_KEY"
{ "name": "get_opposition_brief", "arguments": { "applicationNumber": "1712345", "ownApplicationNumber": "1698765" } }
Brief to respond to an opposition
GET /reports/opposition-response-brief?applicationNumber=… — permission reports:read or
reports:opposition-response-brief:read. MCP: get_opposition_response_brief.
What you need to respond to the oppositions your mark received, in one call.
| parameter | what it is |
|---|---|
applicationNumber |
required — the attacked mark |
limit |
cap of each list (opponents, edges, marks): default 10, max 20 |
| field | what it is |
|---|---|
trademark |
the same card as in the filing brief |
oppositionActs |
the opposition acts on the mark, collapsed by fact (versionsOfFact), oldest first, up to 50 + hasMore: date, section with sectionLabel in three languages, the act's citedNorms, its resolved opponents (by id) and unresolvedOpponentMentions |
opponents |
each resolved opponent (most acts first): actsOnTrademark, oppositionsFiled (its opposed edges) and trademarksInClasses — its current marks in the attacked mark's classes |
unresolvedOpponentMentions |
the sum of the unresolved mentions |
coverageFrom, dataAsOf, coverageThrough |
as above |
⛔ An opponent's marks are found by its persona_id, never by its name: by name, "ACME" would drag in
"ACME CHILE"'s. ⛔ An opponent mention that matches no single party is counted and not named:
its text does not appear in the response. No verdicts and no closing dates outside milestones either.
Unknown mark: 404 not_found.
curl -sS "https://api.tarno.cl/reports/opposition-response-brief?applicationNumber=1712345" \
-H "X-API-Key: $TARNO_API_KEY"
{ "name": "get_opposition_response_brief", "arguments": { "applicationNumber": "1712345" } }
Facts of a filing
GET /reports/filing-context?type=…&applicationNumber=… — permission reports:read or
reports:filing-context:read. MCP: get_filing_context.
The facts a filing about a trademark needs, each with its source, in one call. ⛔ The core gives the facts; the drafting is yours. No draft, no legal prose, no arguments, no LLM: drafting, per-organisation templates and the final document belong to the consumer.
| parameter | what it is |
|---|---|
type |
required — power_of_attorney, opposition, opposition_response (answering an opposition), office_action_response (answering INAPI's observations), withdrawal |
applicationNumber |
required — the trademark of the filing (the target of an opposition; the mark that received the opposition or the observation in a response) |
annotationId |
the act the filing answers: required with office_action_response, optional with opposition_response (one opposition act instead of all). It must be an act of that mark (otherwise 404); with the other types, 400 |
ownApplicationNumber |
only with opposition: your trademark, to add the comparison with the target; with the other types, 400 |
expand=legalTexts |
the text of the cited articles (see below); also requires legal:read |
| type | what facts it brings |
|---|---|
power_of_attorney |
the mark, its classes, its current holders and current representative, each with its tax id (RUT) |
opposition |
the target (publication, classes, status, parties), whether it is today in its opposition window (membership), the oppositions it already received (each act with its date), the opposition history of its holders, the norms cited in those acts, the norm signals of the opposition acts received by its holders' marks (normSignals) and, with your mark, the channel-by-channel comparison (comparison, the one of compare_trademarks) |
opposition_response |
the opposition acts on the mark (or the one you name), their resolved opponents with their tax id, the unresolved mentions counted and the cited norms |
office_action_response |
the act (act.section, act.sectionName, act.date) and the norms it cites |
withdrawal |
the mark, its holder, its published status and its classes |
| field | what it is |
|---|---|
trademark |
the mark's card (with its milestones, publishedOn verbatim and its current parties with taxId) |
facts[] |
{key, value, source} — source {kind, ref, date} says where the fact comes from (trademark, party, annotation, opposition_brief, opposition_response_brief) and what to ask for to read it again (ref: the application number, the party id or the annotation id). ⛔ A fact without a value is omitted, never filled |
milestones |
the terms of the filing with their norm and article: the opposition close (opposition) and the answer to the transfer (opposition_response) |
milestonesOmitted[] |
the terms that are not given, and why (kind, reason, section) |
citedNorms[] |
the norms cited in the acts of the filing, each with its act (source, section); citedNormsTruncated when more than 20 acts cite norms |
normSignals[] |
only with opposition: norms of the opposition acts received by the holders' marks, with counts and evidence — signals, not grounds |
comparison |
only with opposition and ownApplicationNumber: your mark (A) against the target (B); null otherwise. Also null when either has no comparable denomination (a figurative mark, only signs): the file still comes, with a comparison.omitted fact (value no_comparable_denomination, source.ref = that mark's application number) |
outline |
a generic structure (generic: true): the sections header, parties, facts, grounds, requests and signature, and which fact keys (or fields) fill each one. No text and nobody's brand |
coverageFrom, dataAsOf, coverageThrough |
since when the corpus has oppositions, and freshness |
Terms: only with their article. The file publishes two terms, the SAME milestones the trademark
publishes. With opposition, the opposition close (opposition_close): 30 business days from the
publication of the extract, with its legal basis in the field (Ley 19.039 art. 5 and art. 11; Ley 19.880
art. 25). Without a publication date there is no anchor: it is omitted with no_publication_date.
⭐ With opposition_response, the answer to the transfer (opposition_response_close, since
2026-10-10): the 30th business day after the transfer was notified, not extendable (Ley 19.039 art. 9 and
art. 11; Ley 19.880 art. 25; Decreto 82/2022 art. 24). Not answering is not abandonment: once the term
has elapsed the procedure goes on. The anchor carries the act the date comes from (annotationId,
sectionCode), which also travels as the opposition.transfer fact. It is derived when the corpus has
the Estado Diario notification (C1) and the transfer resolution on the same day, and estimated —
say so— when one of the two is missing (estimateReasons). For example, the transfer of file 1605530,
notified on 2025-05-22, gives 2025-07-04, the same date INAPI recorded as «Fin de plazo». When it is not
given, milestonesOmitted says why: no_transfer_on_record (the corpus does not have the transfer yet:
it arrives weeks after the opposition), madrid_designation (INAPI sets another term for a Madrid
designation), prior_legal_regime (a transfer before 2022-05-09, notified by registered letter),
anchor_on_non_business_day or outside_holiday_calendar. It means «not known», never «no term».
The terms that have no article are not given: the answer to an office action on a filing
(office_action_response_term, with its section: 661B, 813B, 931, 925) is listed in
milestonesOmitted with reason: "no_legal_basis_in_knowledge_base" — each resolution sets them and they
can be extended. ⛔ Do not compute or show a date for them as if it were a term from the core.
expand=legalTexts adds to each citedNorms[] the literal text of the article/letter that applied
on the act's date (text, textStatus) and the INAPI documents that cite it (guidance, signals with
counts): the same explained citation as get_annotation?expand=citations. It
requires legal:read as well: without it, 403 forbidden naming legal:read, and no quota is
consumed. Without expand those three keys are absent.
There is no type for third-party observations. The INAPI knowledge base was checked and
it does not record them as a filing distinct from the opposition — there, «observations» are INAPI
resolutions about a filing (661B, 813B…), which office_action_response already covers. Asking for
type=third_party_observations is a 400. If the knowledge base comes to record them with their norm,
they will be a sixth type.
Own permission: neither an unrestricted key, nor the trial key, nor the opposition briefs' permissions
(reports:opposition-brief:read, reports:opposition-response-brief:read) get it. Without it, 403
forbidden naming reports:filing-context:read, with no quota consumed. Nonexistent mark: 404.
curl -sS "https://api.tarno.cl/reports/filing-context?type=opposition&applicationNumber=1712345&ownApplicationNumber=1698765" \
-H "X-API-Key: $TARNO_API_KEY"
{ "name": "get_filing_context", "arguments": { "type": "office_action_response", "applicationNumber": "1712345", "annotationId": "84213", "expand": ["legalTexts"] } }
Scopes
All four reads require the brands:read scope. An empty scope array is unrestricted for them (same
rule as the rest of the contract).
The reports have their own permission: reports:read grants all of them, and
reports:<report>:read one (reports:daily-bulletin:read, reports:clearance:read,
reports:portfolio:read, reports:opposition-brief:read,
reports:opposition-response-brief:read, reports:conflict-landscape:read,
reports:registration-signals:read, reports:filing-context:read). ⛔ They are granted
explicitly: a key with an empty scope array does not get them, nor does the trial key. Without the
permission, 403 forbidden. Each report call counts one unit of quota. See the
authentication guide.
MCP
REST is equivalent to MCP: same key, same scopes, same schemas.
| MCP tool | REST equivalent |
|---|---|
get_trademark (with expand) |
GET /trademarks/{applicationNumber}?expand=… |
list_trademark_annotations (family=opposition) |
GET /trademarks/{applicationNumber}/annotations?family=opposition |
search_trademarks (oppositionWindow) |
POST /trademarks/search |
get_renewal_calendar |
POST /trademarks/renewal-calendar (without portfolio, also GET) |
get_daily_bulletin |
GET /reports/daily-bulletin |
get_clearance_report |
GET /reports/clearance |
get_conflict_landscape |
GET /reports/conflict-landscape |
get_registration_signals |
GET /reports/registration-signals |
get_portfolio_report |
GET /reports/portfolio |
get_opposition_brief |
GET /reports/opposition-brief |
get_opposition_response_brief |
GET /reports/opposition-response-brief |
get_filing_context |
GET /reports/filing-context |
Errors
All errors use the RFC 9457 body described in getting started.
| HTTP | code |
When |
|---|---|---|
| 401 | unauthorized |
Key absent, malformed, unknown or revoked. |
| 403 | forbidden |
Missing the brands:read scope, or the report's permission (reports:read or its own), or legal:read with expand=legalTexts. |
| 404 | not_found |
Nonexistent mark (also a brief's target or own mark), or nonexistent party in the portfolio; in the filing facts, an annotationId that is not an act of that mark. |
| 400 | invalid_rut / invalid_range |
A taxId not shaped like a RUT, or a range with From > To — also with similarity and the opposition window. |
| 400 | validation_error |
Missing required parameters, an unknown parameter, similarTo with similarToTrademark, the opposition window combined with similarity, or a renewal calendar without from/to or longer than 366 days. |
| 422 | out_of_coverage |
Bulletin of a day with no ingested bulletin, or an opposition window with a holiday calendar that does not cover today − 90 to today + 60 days (carries coverage). |
See also
- Authentication and scopes —
X-API-Key,brands:read. - Connect your AI agent (MCP) — the same reads via MCP.
- Getting started — envelope, keyset pagination, error body.
- The live OpenAPI reference at
https://api.tarno.cl/docs.