Análisis competitivo · cada fila con su comprobación
6 ejes.
2 que no sostiene ningún proveedor establecido.
3 que no sostenemos nosotros.
Decenas de proveedores venden trozos de esto. El informe fundacional comparó a catorce de ellos en seis capacidades y concluyó que ningún competidor sostenía más de tres. Volver a comprobar esa matriz la cambió. Cinco de las seis capacidades originales no las podía comprobar un lector, así que se eliminaron. Se añadieron cinco comprobables. 3 de las seis que sobrevivieron son ejes en los que ORBIS pierde, y se retiraron 12 afirmaciones distintas porque se habían resuelto a nuestro favor.
Servidos por la celda que ha generado esta página. Ábralos todos si quiere.
LIVE/conformance/vectorsLIVE/openapi.jsonLIVE/.well-known/did.jsonLIVE/capabilitiesThe matrix
Six axes, six vendorsThe rule that built this table: every axis carries a check a reader can run. An axis that could not be justified was deleted rather than softened — a short matrix that survives scrutiny is worth more than a long one that does not.
| Axis | ORBIS.ID this cell | Microsoft Entra Verified ID a service on the Entra tenant | MATTR credential platform, NZ | walt.id open-source kit | Procivis One Orell Füssli, CH | SpruceID US state mDL contractor |
|---|---|---|---|---|---|---|
| Publishes its OWN test vectors, with a checksum manifest anyone can recompute | HOLDS | no | no | no | no | no |
| A stranger can check a credential with no account, no key and no contract | HOLDS | no | no | no | no | no |
| Serves public, unauthenticated discovery documents on its primary host | HOLDS | unmeasured | partial | unmeasured | no | no |
| Issues in more than one credential format | WE DO NOT | no | holds | holds | holds | unmeasured |
| Holds an independent security certification or a published third-party code audit | WE DO NOT | holds | holds | unmeasured | unmeasured | unmeasured |
| Runs in a live public deployment at state or national scale | WE DO NOT | holds | holds | unmeasured | partial | holds |
Ours, and provable against the machine that served you this page.
The row names the endpoint or the command. Run it.
Ours, and we do not hold it. 3 of the 6 rows are this.
Every one of them is restated in the gaps below, in plainer words.
Theirs — a reading of that vendor's own public documentation, dated August 2026.
A reading is not a measurement. We may have read it wrong, and if we have, we would rather be told than be right by default.
We could not establish it, and we are not going to guess.
This never means "no". Collapsing the two would be the most attackable thing on this page.
Every axis, and how to check it
Including the ones we loseFive of the founding brief's six capabilities were deleted here, and the reasons are worth stating plainly: key custody and messaging are not served from this cell at all, "issue in under an hour" cannot be timed by anyone outside, "daily-life surfaces" is a judgement rather than a measurement, and our own self-hosting cannot be verified by a reader while the platform repository is private. What is left is what a stranger can establish without taking our word for anything.
Publishes its OWN test vectors, with a checksum manifest anyone can recompute
How you check it: Fetch the vector set, canonicalise it with keys sorted at every depth, take the SHA-256, base64url it, and compare with the digest the set publishes about itself. The recomputation below was done in Python against a TypeScript implementation and matched exactly.
Evidence: /conformance/vectors
Why this axis is here: This axis is narrow on purpose, and the two obvious rebuttals are answered here rather than left for someone else to make. FIRST: machine-checkable conformance testing plainly exists — the OpenID Foundation publishes a free, open-source, machine-runnable conformance suite that vendors wire into CI, and it reports a high pass rate across dozens of wallet and verifier pairs. This axis is not about that. It is about a vendor publishing ITS OWN vectors, at its own URL, with a digest that pins the set. SECOND: the axis is limited to ESTABLISHED vendors, because small projects created recently do publish vector sets, so the unqualified version of this claim is false and is not made. Around 45 vendor repositories were searched. SpruceID's vc-test-suite fork and its vp_interop repository are both archived; walt.id's live ETSI plugtest files publish type metadata that omits the SD-JWT VC integrity checksum field entirely; MATTR and Procivis publish no conformance repository.
A stranger can check a credential with no account, no key and no contract
How you check it: Count the operations in the published API description that carry no security requirement. One jq expression does it, and it is printed below with the number it returned.
Evidence: /openapi.json
Why this axis is here: This is the one column of the founding brief's six that survived the checkability rule, and it is checkable on both sides: ours by the command below, theirs by their own price list. The axis is NOT about price, and framing it that way would be easy to knock down — several vendors here have generous free tiers, and this page deliberately publishes no figure for any of them because we could not verify one first-hand. The axis is about whether an unaffiliated third party can verify AT ALL without an account, and on every rail compared here they cannot.
Serves public, unauthenticated discovery documents on its primary host
How you check it: Six well-known paths, one host, one curl each. The verdict is HTTP 200 AND a body that parses as JSON. The exact command is printed below so the measurement can be repeated.
Evidence: /.well-known/did.json
Why this axis is here: This axis measures PUBLICATION, not capability, and the page says so beside the table. Most of this field is enterprise-deployed with no public instance — Entra issues per tenant — so a vendor answering nothing here may hold the capability entirely. Marks of `unmeasured` are not marks against anyone.
Issues in more than one credential format
How you check it: Read each vendor's own format documentation, and read ours: this cell issues SD-JWT VC and nothing else. The capability register at /capabilities says the same thing.
Evidence: /capabilities
Why this axis is here: An axis ORBIS loses, and it is on the page for that reason. Ranked by documented generally-available breadth, PROCIVIS IS AHEAD OF EVERYONE HERE, including us: SD-JWT VC (with post-quantum ML-DSA-65), ISO 18013-5 mdoc, ISO 18013-7 over BLE and MQTT, W3C VC Data Integrity, BBS+ selective disclosure and VC-JOSE-COSE, none of it labelled beta or preview. Then walt.id, then MATTR. Two things cut the other way and both are documented rather than asserted: Microsoft Entra Verified ID is NARROWER than us in kind — W3C VCDM 1.1 secured as JWT-VC and nothing else, with no mdoc, no ISO 18013, no SD-JWT and no BBS anywhere in its live documentation — and MATTR has narrowed rather than broadened, its current lane listing mDocs and CWT with SD-JWT VC at limited availability in pre-production, issuance only. SpruceID is marked unmeasured because its documentation site is offline and every format claim is source-code-only today. Nobody in this field documents AnonCreds — that gap is the industry's, not ours.
Holds an independent security certification or a published third-party code audit
How you check it: Ask each vendor for the certificate, or read their trust page. Ours would be an empty answer: ORBIS holds none of these today.
Evidence: we hold no security certification and have commissioned no published code audit
Why this axis is here: An axis ORBIS loses outright. MATTR is the clearest counter-example and its marketing description in the source brief — "best-certified" — was itself unsupportable and has been withdrawn below; what MATTR actually holds is listed there and it is substantial.
Runs in a live public deployment at state or national scale
How you check it: Read the issuing authority's own published figures rather than the vendor's. Ours is zero: no member of the public holds an ORBIS credential.
Evidence: no public deployment; no credential has been issued to a member of the public
Why this axis is here: The axis ORBIS loses by the widest margin, and the one a buyer should weigh most heavily. Each figure beside it is attributed to whoever published it and dated, because the two biggest numbers in this field describe different things: California's Governor's office reported 1.7 million ACTIVE mobile driving licences in April 2026, and SpruceID states 4 million ISSUED as of August 2026. Four months and two different quantities separate those, and neither refutes the other. Three disciplines govern the quoting, and each exists because we got it wrong first: a CREDENTIAL count is never reported as a PEOPLE count; "applied", "issued" and "active" are never divided into one another to manufacture a conversion rate, because we hold no source that reconciles them; and where a national scheme is cited the denominator and the word "active" travel with the percentage, or the percentage is not printed at all — which is why no adoption percentage for Denmark or Belgium appears anywhere on this page despite both running against us. We could not open those sources ourselves, so we publish the direction and not the digits.
The discovery probe
Method, command, and its two defectsSix well-known discovery paths, one request each, run from a single vantage point on 19 August 2026. The verdict is HTTP 200 and a body that parses as JSON — both halves matter, and the reason is the first defect below.
$ for h in orbis.id mattr.global walt.id procivis.ch; do for p in /.well-known/did.json /.well-known/did-configuration.json /.well-known/openid-credential-issuer /.well-known/oauth-authorization-server /.well-known/jwt-vc-issuer /.well-known/openid-configuration; do c=$(curl -s -4 -L --max-time 12 -o /tmp/b -w '%{http_code}' "https://$h$p"); [ "$c" = 200 ] && jq -e . /tmp/b >/dev/null 2>&1 && echo "HIT $h$p"; done; doneHIT orbis.id/.well-known/did.json HIT orbis.id/.well-known/openid-credential-issuer HIT orbis.id/.well-known/oauth-authorization-server HIT mattr.global/.well-known/did.json
did-configuration.json scored a hit while
answering 404. The verdict now requires the 200 as well. Second: one host returned
200 text/html on all six paths — a single-page-application fallback that would
have produced six false positives for a verdict that only checked whether a response came
back. A third defect was found earlier still: an IPv6 resolver returned an identical bogus
AAAA record for unrelated hosts, making them appear to agree. Every request is now
curl -4.3 / 6orbis.id
did.json, openid-credential-issuer and oauth-authorization-server answer. did-configuration.json, jwt-vc-issuer and openid-configuration are 404.
1 / 6mattr.global
did.json answers. The other five are 404.
0 / 6entra.microsoft.com
403 on all six — UNMEASURED, not absent. The host refused the probe; it said nothing about the product. Entra issues per tenant, so there is no public instance to measure.
0 / 6walt.id
The apex answers 404 with a JSON body on all six. The service subdomains (issuer.portal, verifier.portal, wallet) returned connection errors on three consecutive runs from this vantage point — UNMEASURED, not absent.
0 / 6procivis.ch
404 on all six.
0 / 6okta.com
404 on all six.
0 / 6pingidentity.com
404 on all six.
0 / 6truvera.io
404 on all six.
0 / 6trinsic.id
404 on all six.
0 / 6yoti.com
404 on all six.
0 / 6world.org
404 on all six.
0 / 6signicat.com
404 on all six.
0 / 6onfido.com
403 on all six — UNMEASURED.
0 / 6bankid.com
200 with text/html on all six — a single-page-application fallback. Under a verdict of "the body came back" this host would have scored six false positives.
0 / 6mitid.dk
404 on all six.
0 / 6account.gov.uk
HTML on all six — UNMEASURED by the JSON verdict.
What can be proved right now
Four commands, four answersEverything in this section was run against orbis.id on
19 August 2026 before it was written down. Two of the four answers are bad for us, and they
are here for the same reason the other two are.
The conformance vectors, and a digest anyone can recompute
The published set is 48,988 bytes behind no authentication. Its
digest is sha256 of the canonical JSON of the vector array, base64url-encoded,
and the set publishes that digest about itself. An independent recomputation in Python,
against the TypeScript implementation that minted it, matched exactly. The digest is the
proof; the adjective is not.
$ curl -s https://orbis.id/conformance/vectors -o v.json && python3 -c "
import json,hashlib,base64
d=json.load(open('v.json'))
c=json.dumps(d['vectors'],sort_keys=True,separators=(',',':'),ensure_ascii=False)
print(base64.urlsafe_b64encode(hashlib.sha256(c.encode()).digest()).decode().rstrip('='))
print(d['digest'])"eBNGkInlcF9M7sQTV5KWA9HMP-lrdJr-dvyL-kkkPOU eBNGkInlcF9M7sQTV5KWA9HMP-lrdJr-dvyL-kkkPOU # Recomputed in Python against the TypeScript implementation # that minted it. Measured against orbis.id on 2026-08-19.
orbis.id, and that is deliberate. The cell serving you this page serves
16 vectors at
/conformance/vectors, not the production set. Both numbers
are true about different machines, and a page that printed the larger one while pointing a
link at the smaller would be doing the exact thing this site exists not to do.Verification with no account, no key and no contract
One expression counts the operations that carry no security requirement at all. There is nothing to sign up for and nothing to agree to.
$ curl -s https://orbis.id/openapi.json | jq '[.paths[][] | select((.security // []) | length == 0)] | length'
57 # of 271 operations in total, across 251 distinct paths. # Measured against orbis.id on 2026-08-19.
Microsoft's own documentation is the clearest contrast, and it is worth quoting rather than characterising. From the Entra Verified ID pricing page: “Verified ID generates individual billing events for each unique verification performed by the platform, whether that verification succeeds or fails.” Their own billing matrix on the same page makes it concrete — a verification request that returns a failed face match still emits a billing event.
Two things must be said with that, because leaving either off would make our side look better than it is. Microsoft's Azure pricing page offers up to 50,000 transactions per month at no cost, so this is not an argument that Microsoft is expensive. And Face Check is a separately billed meter that sits outside that free tier, so the quote above and the free allowance describe two different things. The distinction that actually survives is neither of those: it is that an unaffiliated third party needs an account before they can check anything at all.
Revocation, which has never once been used
This one is against us and it is the most credible thing on the page, because anyone can run it.
$ curl -s https://orbis.id/status/1 | cut -d. -f2 | base64 -d 2>/dev/null | jq -r .status_list.lst | base64 -d | gunzip | xxd -b | awk '{for(i=2;i<=7;i++) s+=gsub(/1/,"1",$i)} END {print s" bits set"}'0 bits set # 512 bytes, 4096 single-bit entries, none of them set. # Nothing has ever been revoked.
What we withdrew
14 claims, 12 against usEvery row below is a claim the founding brief made, what re-checking it found, and what this site publishes now. 12 move against us. 2 move the other way, and both of them make an incumbent look larger than we had said. Publishing a competitive page while knowing these were wrong was not an option; publishing them quietly was not either.
Jumio pricing — withdrawn twice, because the first withdrawal overreached
It stood as: $3.00–8.00 per check, described as a published rate.
Re-checking found: Withdrawn three times, and the third withdrawal is the honest one. FIRST we called it a published rate; jumio.com/pricing returns HTTP 404 — fetched for this page on 19 August 2026 — so it is not. SECOND we over-corrected to "Jumio publishes no price at all", which is a claim we cannot support either. THIRD we were told a specific public per-verification price existed on a cloud marketplace listing; when we went to look for it ourselves it was not visible, so we will not print it.
What we publish now: One sentence, and it is the only one we have opened a page to support: jumio.com/pricing returns 404, checked 19 August 2026. We make no claim about what Jumio charges, in either direction. The $3.00–8.00 band is withdrawn outright — it came from a third-party estimate aggregator whose method and sample we never verified, and we published it as though it were Jumio's own rate. Using a competitor's unverifiable price to make our free verification look better was the single most attackable thing on this page.
Okta credentials
It stood as: Not shipped.
Re-checking found: Okta shipped Digital ID Verification as a limited beta on 12 January 2026, for non-production and test environments — inside the Early Access window Okta had itself announced. The first correction we wrote for this row was also wrong: it said general availability was targeted for FY27, which belongs to a DIFFERENT product.
What we publish now: Okta shipped Digital ID Verification as a limited beta on 12 January 2026, for non-production and test environments. Its wider Verifiable Digital Credentials platform is targeted at FY27 (February 2026 to January 2027). Okta never announced FY27 general availability for Digital ID Verification, and we should not have written that it did.
MATTR certification
It stood as: "Best-certified cohort member."
Re-checking found: Unsupportable as written — MATTR is absent from both OpenID Foundation verifiable-credential registers. What it does hold is substantial and organisational: SOC 2 (its page says "attested" and does not state Type I or Type II), ISO 27001 and ISO 27701, and a Trail of Bits code audit that Trail of Bits' own index dates to May 2022 with no published report.
What we publish now: MATTR holds an ISO 27001 and ISO 27701 certification set plus a SOC 2 attestation, and commissioned a code audit. That is more than ORBIS holds and we say so. The distinction we draw is narrow and it is the only one that survives: those are ORGANISATIONAL assurances, not protocol conformance, and none of them is checkable from a public register the way a conformance listing is. We do not assert SOC 2 Type II, because MATTR does not. And the claim is "has not certified" — never "has not tested": MATTR took part in OpenID Foundation interoperability testing in July 2025.
California mDL volume — a correction we had to withdraw AGAIN
It stood as: SpruceID — CA DMV, 4M+ issued, read as a vendor number to be discounted.
Re-checking found: Three attempts, and the first two were both wrong. FIRST we accused SpruceID of inflating California's numbers; they had not. SECOND we replaced the accusation with a ratio derived from a California DMV analytics dashboard — and that was worse, because the figures were not ours to quote. THIRD, and this is the correction to our own correction: we wrote that the dashboard host did not resolve. It does. Measured here on 19 August 2026, analytics.dmv.ca.gov resolves and answers HTTP 200 with 1,274 bytes of a client-rendered application shell — the figures load in the browser afterwards and are simply not in the page the server returns, so nobody here could verify them from what it serves. That is the honest finding, and it fully justifies the strike; "the host does not resolve" was a stronger claim that happened to be false, inside the very paragraph admitting a false claim. What two people on this side have now opened and read for themselves is this: the Governor's office reported 1.7 million ACTIVE mDLs on 28 April 2026, approximately 900,000 of them in the DMV's own wallet app; and SpruceID's own page states 4 million mDLs ISSUED as of August 2026. Those two numbers are four months apart and count different things. Neither refutes the other.
What we publish now: California's Governor's office reported 1.7 million active mDLs in April 2026, approximately 900,000 of them in the DMV's own wallet app. SpruceID, which built that app for the DMV, states that 4 million mDLs had been issued programme-wide as of August 2026 — across the DMV, Apple, Google and Samsung wallets. Each number is attributed to whoever published it and dated, and NOTHING is computed from them: applied, issued and active are three different quantities and we hold no source that reconciles them, so no ratio, funnel or conversion rate appears here. Two disciplines survive from this row and govern the whole page: a credential count is never reported as a people count, so "X million Californians" is never written; and no number is published unless someone on this side opened the page it came from and can quote the string.
Truvera per-verification price
It stood as: $2.00 per verification, from $499/mo divided by 250.
Re-checking found: The unit is per credential ISSUED. Truvera does not meter verification at all.
What we publish now: Withdrawn, and not replaced. The cell compared a price for one thing against a price for another, which made our free-verification argument look stronger than it is. We publish no Truvera price at all, having opened no Truvera pricing page ourselves.
Yoti fine
It stood as: Fined €950k, March 2026.
Re-checking found: The two qualifiers we were given — that it did not concern age verification, and that it is under appeal — are material, and neither we nor anyone in this chain opened a regulator page to confirm the penalty, its subject or its status.
What we publish now: Nothing. The row is withdrawn entirely rather than restated with qualifiers we have not verified. A competitor's regulatory penalty is the last thing to publish on someone else's say-so.
World / Worldcoin
It stood as: 17.5M orb scans; banned or halted in 12 jurisdictions.
Re-checking found: Both halves were wrong in our favour: the scan count was not their own wording, and the count of jurisdictions was higher than anything that can be enumerated. We did not open their page or a regulator's to fix either.
What we publish now: No figure. Neither the verification count nor the restriction count is published here, because we have not verified either first-hand, and an unverified count of a rival's legal troubles is the worst kind of number to guess at.
Microsoft Entra Verified ID
It stood as: "A feature of the Entra tenant", implying it must be bought.
Re-checking found: We replaced it with "free to 50,000 transactions a month" — and could not stand that figure up when we went to check it ourselves, so it is withdrawn too. What we DID fetch and can quote is Microsoft's own Face Check billing article, and it cuts our way while also qualifying itself: two of the four scenarios in Microsoft's own matrix emit NO billing event.
What we publish now: We make no claim about what Microsoft costs. The distinction that survives has nothing to do with price: it is whether an unaffiliated third party can verify at all without an account, and on Microsoft's rail they cannot.
walt.id funding
It stood as: Pre-seed funded.
Re-checking found: The description traces to a funding round several years old, though we did not open the filing or announcement ourselves to date it precisely.
What we publish now: Dropped, with no replacement. A funding stage presented as current status is not a fact about today, and we have not verified which stage applies now.
Procivis engagements
It stood as: "Lithuania sandbox, WE BUILD pilot" — read as one engagement.
Re-checking found: Two separate engagements. Lithuania serves the Aptitude consortium, and Procivis won it by public procurement.
What we publish now: Listed separately, with the procurement noted. We were understating a competitor.
Swiss swiyu
It stood as: Listed as a Procivis credential.
Re-checking found: swiyu is built by the Swiss federal government (FOITT and the Federal Office of Justice) and is open source. Procivis is interoperability-TESTED against it.
What we publish now: Compatibility, not a contract. The two are not the same claim.
Trinsic — two errors in one line
It stood as: Sold its SSI platform, buyer recorded as Jumio. Read as a company leaving the category.
Re-checking found: Two things were wrong. The buyer was Dentity, announced 12 December 2024. And Trinsic was not acquired at all — it DIVESTED its decentralised-identity platform and still operates today under the same chief executive. Jumio and Trinsic announced a commercial integration on 25 June 2026, which is more recent than the divestiture we were citing.
What we publish now: Trinsic sold its decentralised-identity platform to Dentity in December 2024 and pivoted to its Identity Acceptance Network. We have dropped the "not Jumio" phrasing as well: it is technically true and reads as uninformed to anyone who follows the market.
Denmark — the scheme was named wrong, and the number was too low
It stood as: "Denmark's BankID", 93% of Danes.
Re-checking found: BankID is Swedish and Norwegian; Denmark's scheme is MitID. That part is a plain naming error and it stands corrected. The replacement PERCENTAGE, however, is a figure nobody on this side has opened a page to confirm — the attempt to re-fetch it for this page failed to connect — so it is not printed here.
What we publish now: Denmark's national scheme is MitID, not BankID. We publish no adoption percentage for it, because we could not verify one first-hand — and the direction it would move in is against us, which is precisely why we are not going to reach for an unverified version of it. The qualitative point survives and is not in doubt: MitID's reach among Danish residents is overwhelming, and ORBIS has no deployment at all.
Belgium itsme reach
It stood as: 80%.
Re-checking found: The 80% figure is stale. The replacement user count we were given is not something anyone here opened itsme's own page to confirm.
What we publish now: No figure. The direction is not in doubt and it runs against us — Belgium's national scheme is far larger than we said — but we will not publish a specific number we have not seen with our own eyes, even one that flatters a competitor.
The honest gaps
Where this loses, and where nobody wins4 of these 7 are competitive losses: the field holds them and we do not. 3 are not — they are gaps across the entire measured field, and presenting them as competitive weaknesses would overstate our own weakness, which is its own kind of inaccuracy. The badge on each row says which kind it is.
Pre-revenue, and nobody outside this project uses it
No member of the public holds an ORBIS credential. Every deployment figure on this page belongs to somebody else.
No security certification and no published code audit
ISO 27001, SOC 2 and eIDAS conformity are budgeted and sequenced, and none of them is held today. MATTR holds three of them plus a Trail of Bits audit.
One credential format
SD-JWT VC, which the EU implementing acts cite directly. mdoc and mDL are designed behind the same endpoints and are not built. Four of the five vendors compared here issue in more than one format today.
Domain linkage is not published — and this one is a real loss
/.well-known/did-configuration.json answers 404 on our host. It answers 200 with JSON at DIF, Sumsub, Persona and Talao, all measured the same night as the matrix above. Until it is served, a reader is asked to accept the binding between our domain and our DID on DNS alone.
SD-JWT VC issuer metadata is not published — and this one is NOT a competitive loss
/.well-known/jwt-vc-issuer answers 404 on our host, and it answered 404 on every host in the measured field, DIF and Sphereon included. It is a gap in the whole field. Presenting a universal gap as something the competition has and we lack would overstate our own weakness, which is a way of being wrong about ourselves.
Revocation has never been exercised
The signed status list decodes to 4,096 entries with zero bits set. Nothing has ever been revoked, because nothing has ever needed to be. The machinery is real and untested by use, and anyone can decode the list and confirm both halves of that sentence.
Not certified — stated accurately rather than dramatically
ORBIS holds no verifiable-credential conformance certification. That is the whole of what we will assert, and the reason is worth reading. We were told, twice and in conflicting terms, what the OpenID Foundation's verifiable-credential certification programme currently looks like: one account gave an opening date and three certified organisations, another said no such register was published. So we opened the OpenID Foundation's own certification page ourselves on 19 August 2026. It does not mention verifiable credentials, OpenID4VP, OpenID4VCI or HAIP at all, and lists no certified organisations for any of them. We therefore publish NO date, NO count and NO names — not the flattering version, not the unflattering one. One thing must NOT be inferred from this and is not said anywhere on this site: that established vendors are "not OpenID-certified". Several hold OpenID Foundation certifications on other programmes entirely, and conflating the two would be exactly the kind of cheap shot this page exists to avoid.
The graveyard
Not “we are better” — “they stopped”The founding brief's own section, ported with one correction. It recorded Trinsic as having sold up and left. Trinsic divested its decentralised-identity platform to Dentity in December 2024 and is still trading — so the entry now says what ended, which is the pure-play business, rather than the company.
Ask this site
Try:
- IssuerSigns a claim about you. Learns nothing afterwards.
- Holder — youKeeps it. Chooses what to show, and to whom.
- VerifierChecks the signature and the status. Does not call the issuer.
The arrow that is missing is the one that matters: there is none from the verifier back to the issuer, which is why the issuer cannot see where you went.
- Read the token and the key it names.
- Resolve the issuer’s key document.
- Recompute that key’s thumbprint and compare it to the name in the token.
- Verify the signature, using the algorithm the token itself declares.
Every step is arithmetic the receiving party performs. None of it asks the issuer to vouch for the issuer.
ExampleA credential with four claims, of which two are sent:
| Claim | Value | Sent? |
|---|---|---|
| given_name | Ada | sent |
| date_of_birth | 1970-01-01 | withheld |
| over_18 | true | sent |
| address | … | withheld |
The issuer’s signature still verifies over the two that were sent, because the two that were not are committed to as salted digests rather than as values.
Everything this search can see — all 27 pages, and what each one claims
- ORBIS.IDYour identity belongs to you. Right now, it belongs to everyone else.Every time you sign up for something, you hand over a copy of yourself — a photo of your passport, a scan of your face, your date of birth. Those copies pile upSections: The grievance, the Article, and the proof · The problem, in one sentence you already know · Five promises — and what each one actually means · Three people, one seal, no copies · The screen where you decide · What we can see, and what we cannot
- Declaration of Digital IndependenceUniversal Declaration of Digital Independence and Human Digital RightsThe founding document of this estate, written by Jonatan Schmidt and published here unedited. Next to each Article is what the running system actually does abouSections: Universal Grievances Against Digital Oppression · Universal Principles of Digital Sovereignty · A Universal Call to Unified Action · In Witness Whereof · The honest case against this whole idea
- The Trust FabricTrust, woven in.The trust fabric for the real economy — it turns everyday life into proof you own . An issuer signs a statement, a holder keeps it and decides who sees it, and Sections: Split trust into three roles — never merge them. · A bill becomes proof. Proof becomes credit. · Honesty is the architecture. · The fabric cannot be retrofitted. · Three converging markets — one regulatory clock. · The lines the product already says for itself.
- Demo consoleThe only demo here is the real thing.There is no recorded walkthrough on this page, no simulated data and no screenshot of a product. There is a running trust plane and the commands that talk to itSections: 01 · The trust anchor · 02 · What this issuer issues · 03 · Is it healthy right now · 04 · Revocation, signed · 05 · Prove it against the published conformance vectors · Two lists, same page, same weight
- DocumentationGenerated from the route table, so it cannot drift.The wire contract is derived from the live routing table on every request. The document and the server cannot disagree, because one is made out of the other.Sections: Start here · The full contract
- DevelopersEverything below is a live response.Not documentation of an intended one. Five commands, no account, no API key, no sales call. Run them before you decide whether to keep reading.Sections: 01 · The trust anchor · 02 · Issuer metadata · 03 · The credential shape · 04 · Conformance vectors — check your implementation, not ours · 05 · Revocation, signed · Readiness that reports the hardware
- CapabilitiesEvery claim, its real state, and the proof.25 entries: 14 live · 1 partial · 1 planned · 9 not yet. Each one was checked against the cell serving this page. Almost nobody in this industry publishes the rSections: What the four words mean · Identity · Issuance · Operations · Verification · Standards
- StandardsOne row per standard. Including the ones we fail.5 live · 4 partial · 0 planned · 3 not yet. Every row states what the standard demands , what this cell actually does, and the artefact that settles it. Two rowSections: What the four words mean · Credential formats, as their own register · Why every row carries a state
- RegulationsThe posture, in the same discipline.Conformant where it is true, planned where it is not yet, and never silent about the difference. Data minimisation here is a property of the credential format, Sections: Data minimisation, enforced by cryptography · What the four words mean · Compliance and security capabilities · Security · Compliance · Certification status, stated once and plainly
- For peopleThe answer is smaller than the question.You photographed your passport again this month. Someone needed to know one thing about you. They now hold your name, your face, your address and your date of bSections: Prove the fact. Keep everything else. · What changes for you · What is real today, and what is not · If you are here for someone else
- For businessYour own issuer. Your own domain.You mint your own credential types under your own domain. Your members hold them. Your counterparties check them offline, against a public key, with nothing to Sections: You are not early. The deadlines already exist. · The live example is not a mock · What a buyer actually asks · Two lists, same page, same weight · Getting started
- For citiesA city is a tenant, with its own key.Resident credentials a city issues under its own did:web , checked by services across the city with no shared resident database in the middle — because the checSections: Why the architecture is not city-specific · What a city actually gets from this · Two lists, same page, same weight
- For governmentTwo questions, both answered with endpoints.What does it conform to, and what is provable without taking our word for it. Everything below links the artefact that settles it — a document you fetch, a fileSections: The property that comes before everything else · Conformance, one row per standard · Accountability · Two lists, same page, same weight
- PartnersFour roles, and what each one runs.Issuer partners mint their own types under their own tenant. Relying parties verify against the trust anchor. Wallet partners hold on a person's behalf. Data paSections: A page each, with its own evidence · How intake works, and what it costs you · Two lists, same page, same weight
- Apply as a partnerApply, and here is exactly what happens next.The intake is a public, rate-limited endpoint you can call right now. What happens after it is a person, not a pipeline, and this page says which steps are whicSections: Step 1 · Call the intake · Do it here · What each step actually is · What we do with what you send
- Application statusWhere your application is, without asking us.A public endpoint answers with the state of an application you hold the id for. No account, no sign-in, no email thread.Sections: Ask it directly · Check it here · What the states mean
- Issuer partnersYour own key, your own domain, your own exit.An issuer partner runs a tenant with its own did:web , its own signing key, its own status list and its own credential types. The credentials you mint are verifSections: What you actually get, and you can look at all of it now · The formats, as a register · The exit, stated before you sign anything · Two lists, same page, same weight
- Relying-party partnersVerification takes no key. Count the routes.There is nothing to buy and nothing to sign up for on the verifying side. The verification plane takes no authentication at all, and that is a structural properSections: Count it yourself · What you integrate: three fetches · Test your verifier against ours, offline · What is not proved yet, on the record
- Wallet partnersWhat a wallet has to speak to interoperate.The protocols, the format, and the key custody we expect. And the thing no other page in this industry will tell you: nothing has ever completed this flow againSections: The discovery surface, live now · What we expect of you, and what you can expect of us · Aligned with HAIP 1.0
- Data partnersA signed contract, an audit chain, and a short list.A data partner exchanges verified attributes under a signed agreement, with every exchange recorded in a tamper-evident chain. It is also the least built of theSections: What is real underneath · Compliance · What a data partner would get and give · If you want to be first
- Verify with ORBISThe plane designed to survive everything else being down.Verification needs no signing key, no issuer identity and no vault — a cell can be configured for verification alone. It also needs no permission: 107 of this cSections: Three fetches, no account · The honest line between implemented and witnessed · Two lists, same page, same weight
- The clock — the deadlines, counted liveEvery date on this page was set by a parliament, not by us.This is not a roadmap. It is a list of legal instruments with dates attached, and the first one has already gone. Nothing below is written down and left to rot:Sections: The instruments, in the order they bite · Switzerland, and what it actually tells you · What “certified” currently means in this field · To a government reading this · To the people building the same thing · To the person these deadlines are actually about
- How ORBIS compares — and where it loses6 axes. 2 no established vendor holds. 3 we do not hold ourselves.Dozens of vendors sell pieces of this. The founding brief compared fourteen of them across six capabilities and concluded that no competitor held more than threSections: The matrix · Every axis, and how to check it · The discovery probe · What can be proved right now · What we withdrew · The honest gaps
- Your accountAn account with no password to steal.Not a row in our database with your email on it. An ORBIS.ID account is a key , and the half that matters is designed to be born on your own device and never seSections: What an account actually is · How you get one · What the account can do · How to leave with everything · If you are not here as a person
- For your organizationYour own issuer identity — and your own way out.An organization on ORBIS is not an account on our platform. It is a tenant with its own web identity, its own signing key, its own revocation list and its own cSections: What your organization becomes · What onboarding actually involves · What it does not require · Three pages, one for each question · The other two doors
- The operator back officeThe operator console, and the door that will not open yet.Issuance, revocation, the trust registry, the approvals plane, the audit chain and the compliance desk — 195 operator-authenticated operations , the largest surSections: What the back office does · The door, and why it is shut · How an operator signs in · Count it yourself · The other two doors
- Network operations centreEvery panel on this page is a reading your own browser just took.The trust chain resolved live, the revocation field decoded from the signed list, the audit hash chain recomputed link by link, and this cell’s latency measuredSections: The chain that draws itself as it proves itself. · Tamper-evident, and you are welcome to try. · The finding this page would rather report than hide. · What an operator commands, counted from the wire. · Stated here rather than found later.
The words, in plain language — 11 terms, each naming the specification it glosses
- Verifiable credential
- Also searched as: verifiable credential, vc, credential, digital credential, what is a credential, certificate
- A set of claims about someone, signed by whoever issued them. The signature is what makes it verifiable: whoever receives it can prove it has not been altered and can name the key that signed it, without calling the issuer to ask. It is not a login and not a row in somebody’s database — it is a document the subject holds and presents.
- Gloss of W3C Verifiable Credentials Data Model 2.0. The concrete format on this cell is SD-JWT VC (IETF). This site’s own claim: /capabilities · served by this cell: /manifest
- Issuer, holder, verifier
- Also searched as: issuer, holder, verifier, roles, who is the issuer, difference between issuer and verifier, three roles, relying party
- Three roles, separated on purpose. The ISSUER signs a claim about you. The HOLDER — you — keeps it and decides who sees it. The VERIFIER checks the signature and the status. Because the verifier does not have to call the issuer to do that check, the issuer does not learn where you used the credential. That separation is the entire point of the model; collapse it and you have rebuilt a login provider.
- Gloss of W3C Verifiable Credentials Data Model 2.0, which defines these three roles. This site’s own claim: /trust-fabric
- The trust chain
- Also searched as: trust chain, how do i know, real, genuine, authentic, fake, forged, tampered, how is it verified, verification, verify, verified, signature, signed, trust anchor, check, checked, proof, prove
- Checking a credential is four steps, in order, and every one of them can be done by the party receiving it. Read the token and the key it names. Resolve the issuer’s key document. Recompute that key’s thumbprint and compare it to the name in the token. Verify the signature with the algorithm the token itself declares. Nothing in that sequence requires trusting the issuer’s word about the issuer.
- Gloss of RFC 7638 (JWK thumbprint), RFC 7515 (JWS), W3C DID Core for resolving the key. This site’s own claim: /trust-fabric · served by this cell: /.well-known/did.json · /conformance/vectors
- Selective disclosure
- Also searched as: selective disclosure, sd-jwt, sd jwt, withhold, hide claims, privacy, share less, minimum disclosure, only show my age, zero knowledge
- Handing over some claims and withholding others without invalidating the issuer’s signature. Each claim is committed to in the signed payload as a salted hash rather than as the value itself; you send only the disclosures you choose, and the receiver recomputes the digests of what you actually sent. Withholding a claim is therefore not an act of trust or of policy — the arithmetic simply does not contain it.
- Gloss of IETF SD-JWT and SD-JWT VC. This site’s own claim: /regulations
- Revocation and status
- Also searched as: revocation, revoked, status list, expired, cancel a credential, withdraw, still valid, status
- A credential can be withdrawn after it is issued, so a verifier needs a way to ask whether this one still stands. A status list answers that as one bit per credential in a single compressed, signed document: the verifier fetches the whole list and reads its own bit, so the issuer never learns which credential was being asked about. The privacy property comes from the shape of the answer, not from a promise.
- Gloss of IETF Token Status List. This site’s own claim: /trust-fabric · served by this cell: /t/root/status/1
- DID — decentralized identifier
- Also searched as: did, decentralized identifier, did:web, did web, identifier, who signed it, public key
- An identifier that resolves to a document containing public keys. The method decides how that resolution happens. did:web makes the identifier a domain name and serves the document at a well-known path, so resolving it is an ordinary HTTPS fetch anyone can perform and nobody needs a ledger, a token or a permissioned network to complete.
- Gloss of W3C DID Core 1.0; method did:web. This site’s own claim: /standards · served by this cell: /.well-known/did.json · /t/root/did.json
- How a credential reaches a wallet
- Also searched as: oid4vci, issuance, issue, get a credential, openid4vci, how do i receive, onboarding a credential
- The protocol a wallet uses to collect a credential from an issuer: the wallet discovers what the issuer offers from a well-known metadata document, obtains authorization, and requests the credential over an ordinary OAuth-shaped exchange. Being ordinary is the feature — it means an issuer does not need a bespoke wallet and a wallet does not need a bespoke issuer.
- Gloss of OpenID for Verifiable Credential Issuance (OpenID4VCI). This site’s own claim: /standards · served by this cell: /.well-known/openid-credential-issuer · /.well-known/oauth-authorization-server
- How a credential is presented
- Also searched as: oid4vp, presentation, present, show a credential, openid4vp, dcql, prove something, log in with a credential
- The protocol a verifier uses to ask for credentials and a wallet uses to answer. The verifier states what it needs as a query rather than naming a provider, and the wallet responds with a presentation bound to that specific request — so a captured answer cannot be replayed at a different verifier.
- Gloss of OpenID for Verifiable Presentations (OpenID4VP), with DCQL for the query. This site’s own claim: /standards
- Wallet
- Also searched as: wallet, where is it stored, app, phone, my credentials, hold
- The software the holder keeps credentials in, and the place the holder’s keys are created and stay. The property that matters is not the interface: it is that the private key is generated on the device and never leaves it, because a key held by the provider makes the provider the holder no matter what the screen says.
- Gloss of No single specification; the relevant ones are OpenID4VCI/VP for the protocols and the platform key stores for custody. This site’s own claim: /for-people
- What this site is
- Also searched as: orbis, what is orbis, what does this do, what is this, the platform, product, what do you do, help
- This surface will not summarise the product for you, because a summary is exactly the kind of claim this site refuses to make without a receipt. What it will do is hand you the register: the capability page states what is live, what is partial and what is not built, each one linked to the endpoint that proves it.
- Gloss of Not a specification — a register. This site’s own claim: /capabilities · served by this cell: /openapi.json · /build
- The standards
- Also searched as: standards, specifications, specs, interoperable, eidas, regulation, compliance, conformance
- Verifiable credentials are not one standard but a stack: a data model, a serialization, a protocol to issue, a protocol to present, a way to identify keys and a way to publish status. Interoperability claims are only meaningful per layer, which is why a page that claims the whole stack at once is telling you nothing.
- Gloss of W3C VC 2.0, IETF SD-JWT VC, OpenID4VCI/VP, W3C DID Core, IETF Token Status List. This site’s own claim: /standards · served by this cell: /conformance/vectors