CORTEX es un sistema operativo de IA para empresas: unifica los datos, automatiza la operación recurrente y despliega agentes especializados con memoria compartida que analizan, deciden y ejecutan, manteniendo el control humano sobre todo lo importante.

¿Qué parte de la operación puede automatizar CORTEX?

El objetivo es que la empresa opere entre un 80% y un 90% de su actividad digitalizable recurrente sin intervención humana paso a paso: seguimiento comercial, continuidad de clientes, flujo de datos, informes, contenido y detección temprana de problemas.

¿Cuánto cuesta implementar CORTEX?

La implementación llave en mano parte de 15.000 € en pago único; la infraestructura es propiedad de la empresa. La inversión final depende del modelo de negocio, los procesos, las fuentes de datos, las integraciones y el nivel de autonomía.

¿Cómo empieza un proyecto?

Todo proyecto empieza con una auditoría estratégica AI-native de 290 €, que se reserva online y se descuenta íntegra del proyecto si la empresa decide continuar.

¿Los agentes toman decisiones sin supervisión?

No. CORTEX trabaja con cinco niveles de autonomía configurables y controles humanos en cada decisión crítica; los agentes ejecutan lo recurrente y elevan a dirección lo que importa.

There are free agent registries. Why would I pay for this?

Because they are two different things. A registry gives you existence: your agent appears in a list, with an identifier and a badge. That is real, it is free, and we do not compete there. What we publish is the scope: a signed list of what the agent may do and, above all, what it is explicitly forbidden to do, a graded autonomy level, and a page a third party with no contract understands without knowing what a DID is. On top of that, the operation: 90-day rotation, revocation with effect, an exportable auditable history, and a dashboard where an agency keeps several clients’ agents apart. A free registry does not rotate, does not cascade revocation and has no multi-client dashboard.

Is this an accreditation? Does it help me with the AI Act?

It is not an accreditation, and anyone telling you theirs is is selling you something that does not exist. CORTEX Seal is a verifiable self-declaration: you declare your agent’s scope, we sign it, and anyone can cryptographically check that the declaration is yours, is current and has not been modified. It is designed to provide a verifiable answer to the transparency obligation in Article 50 of Regulation (EU) 2024/1689. We do not say it makes you comply with the AI Act, because the AI Act does not require this credential: the obligation rests with the provider or the deployer.

I already pay for an identity provider. Does this overlap?

No, and one question makes it obvious: does your end client have an account in your identity provider? No. Then they cannot verify anything. Your provider controls what happens inside your perimeter, and it does that well; we produce what can be checked from outside it, by someone with no account with you. Many teams use both and there is no technical conflict: the seal does not touch your tokens or your internal policies.

I am not giving you my model-provider keys.

Correct, and we do not want them. The rule is published: we only hold, and only if you switch it on, credentials that revoke themselves — a GitHub or Slack token, exactly the one your agent already carries. Storing it does not widen its reach by a single point. For model providers and clouds we send an HMAC-signed order to your endpoint and you execute the stop at your end. An organisation admin key is something we neither ask for nor want.

What if you revoke by mistake and take down my production on a Friday?

That is why the function ships with a handbrake: a mandatory dry run before the first real use of each resource, a configurable grace window of up to fifteen minutes with a hold button, mandatory dual approval for anything irreversible, and a dialogue that tells you before execution what can be undone and what cannot. Most importantly: revoking the credential and firing the cascade are decoupled. You can revoke the ID without touching a single resource.

A large cloud will do this for free in six months.

Part of it, probably yes, and we say so openly: registries and credential issuance trend towards zero cost, and we do not build the business on that. What does not trend to zero is multi-client operation, the vocabulary of scope and autonomy, and the verification page as an artefact you show. That is why we charge per active agent and not per credential issued, and why we publish the manifest schema and the A0–A5 scale as an open specification. If this gets standardised tomorrow, we would rather be the reference implementation than the proprietary format left outside.

You are a very small company. What if you disappear?

The answer has three concrete parts. One: the manifest schema and the A0–A5 scale are published as an open specification under CC BY 4.0, so the format does not die with us. Two: the credentials are W3C Verifiable Credentials 2.0, a W3C Recommendation since 15 May 2025, verifiable with any standard library without going through our servers. Three: the audit log export is yours, in JSON, at any time, without asking. What you would lose is the dashboard, the automatic rotation and the operational stop button — and it is honest to say so, because that is exactly what you are paying for.

What actually makes you different?

We are not the only ones doing this. We are the ones who publish it so that your end client can check it without an account in our system, and the ones who publish the price, the schema and the list of what we do not guarantee at the same time. If what you need is a free registry, use one. If what you need is internal control, your identity provider already does it. We are the answer when somebody outside has to be able to check.

What exactly counts as an "active agent"?

An agent in the production environment that currently has at least one valid credential — neither revoked nor expired. We take a daily snapshot at 00:05 UTC and write it into your chained audit log, which you can export and recompute yourself. Sandbox, draft and archived agents never count.

What happens if I exceed my agent limit?

We warn you in-app and by email at 80% and 100%. When you try to promote the agent that would exceed it, the API returns a clear error (plan_limit_reached) with a link to add a block or upgrade. We never deactivate or revoke agents already in production over a limit.

Can I change plan mid-cycle?

Yes, from the dashboard, instantly. Proration is automatic: upgrades charge the difference for the remaining days, downgrades credit your next invoice. No penalty, no minimum term on Builder or Agency.

What happens to my credentials if I stop paying?

Already-issued credentials are not revoked. Revoking them would punish third parties who rely on them and have nothing to do with your invoice. What gets blocked is new issuance and renewal; live credentials expire naturally after 90 days. You get a full quarter to land, and your audit export stays available.

How much do I save paying annually?

Two months — pay ten, get twelve. Builder goes from €588 to €490 a year and Agency from €2,988 to €2,490. Annual also includes assisted migration of five agents, otherwise €490.

How does the 30-day guarantee work?

Email support within 30 days of your first subscription and we refund in full, with no questions and no retention call. On annual we additionally refund unused months pro-rata during the first three months. Professional services already delivered, such as the human manifest review or assisted onboarding, are not refundable.

Do prices include VAT?

No. All prices are shown excluding VAT. At checkout, tax is calculated based on your country and on whether you provide a valid EU VAT number; with a valid one, the reverse charge applies. For consumers, the VAT of the country of residence applies. Cosmolabs OÜ is an Estonian company, registry code 17106582.

Why is there no free plan with production agents?

Because a production seal is a signed statement made with our key, with our name behind it and real responsibility if it is misused. Everything else is free: unlimited sandbox issuance forever, full stop-button simulation and a working verifier. You pay the day the seal goes out into the world.

What about anchoring in a distributed ledger?

It is planned and not available. We do not sell it, we do not promise a date, and it is part of no plan today. When it works in production we will announce it with its price. We prefer this answer to a checkbox you cannot use.

Digital Cortex

CORTEX Seal

A verifiable ID for your AI agents. With a stop button.

Issue a signed credential that states what each agent is allowed to do — and what it is explicitly forbidden to do — with a public page anyone can check. When something goes wrong, revoke it: we stop issuing its tokens instantly and send a signed stop order to your systems, with an auditable record of every attempt.

Start in sandbox See a live ID

Unlimited sandbox, no card. Sandbox seals are stamped SANDBOX — NOT VALID: they are for integrating, not for showing.

Live demo

We won’t tell you. Look at it.

Twenty seconds, five states and no video: manifest, credential, verifier in green, revocation, and the same verifier in red. You can pause it and step through it, and the real agent is right below to check it against.

{

"name": "Argos",

"autonomyMax": "A4",

"skills": ["lead.scout", "outreach.draft", "report.status"],

"prohibited": ["outreach.send", "payment.execute", "contract.sign", "data.delete"]

}

manifestHash = sha256(canonicalJson) → 1a651f27f208fb16…

The forbidden list is required, even when empty.

Step 1 / 5

Manifest

You declare what the agent may do and, above all, what it is forbidden to do. When you publish the version, the system computes the document’s SHA-256 hash.

Full sequence: 20 seconds.

Argos is a commercial agent in production carrying credential CRD-2026-0001, autonomy level A4 and four explicitly forbidden actions. Its seal and its checks load here from the same public endpoints any third party would use, with no account and no contract.

CORTEX Seal of the agent Argos, generated from its signed credential.

CRD-2026-0001 · A4

Credential valid · 7/7

Token format
Key declared by the issuer
ES256 signature
Issuer
Validity
Revocation status
Manifest unchanged

Open the full verifier

/functions/agentSeal?slug=argos · /functions/credentialVerify?slug=argos

Declared scope, not just identity

We don’t say who your agent is — we say what you authorised it to do, what you explicitly forbade, and which autonomy level it runs at, from A0 to A5. Signed with ES256 and bound by SHA-256 hash to one exact version of the manifest. If the scope changes, the hash changes and the credential stops matching.

A public page your client actually understands

Eight live checks in plain language, with nobody having to install anything, sign up, or sign a contract with you. It is the artefact you put in a proposal, and the one your end client can open on a Sunday without asking you anything.

Revocation does something

Revoking publishes status to a signed status list, stops token issuance within 5 s, sends an HMAC-signed stop order to your systems within 10 s with retries and acknowledgement, and records every attempt with its response code and latency. What happens inside third-party systems depends on those third parties: we log it and we show it.

How it works

Six steps and one red button.

Step 1
Workspace and key
You create the workspace, choose its identifier — it enters the DID, which is why it cannot be changed later — and receive a sandbox key, shown once.
Step 2
Manifest
You declare the agent: function, tools, permitted scope, forbidden actions and maximum autonomy level. You publish the version and the system computes its SHA-256 hash.
Step 3
Credential
You issue. The private key never leaves the managed key service: the function asks for the signature and receives the token. 90-day validity, with automatic rotation.
Step 4
Verifier and seal
You get a public URL and an SVG seal. Anyone can check format, key, signature, issuer, validity, revocation, manifest and DID document without asking you.
Step 5
Linked resources
You tie the credential to what the agent actually uses: a provider key, a tool server, an endpoint of yours. Each one passes a mandatory dry run before the first real stop.
Step 6
Revoke, with evidence
A mandatory reason, an optional grace window with a hold button, and dual approval if anything is irreversible. The public list reflects it within 60 s and every attempt is recorded with its response code and latency.

What we promise and what we do not

Four numbers, all ours, all measurable on our own systems. And five things nobody can promise, which go in the documentation and not in the small print.

≤5 s

We stop issuing tokens for that agent after revocation.

≤60 s

The public status list reflects the revocation.

≤10 s

Delivery attempt of the signed order, with retries and acknowledgement.

100%

Of attempts are recorded with response code and latency.

What we do not promise

That a key already loaded into the agent’s memory or environment stops working before the provider withdraws it.
That a third party executes the stop, or how long it takes: no provider documents its propagation latency.
That an agent running on infrastructure we cannot reach is stopped.
That a self-contained token dies before it expires on a server that does not check status.
That data already out is recovered, or that an action already executed is undone.

This limitation is declared inside the signed credential itself, not in an annex. Anyone verifying the credential reads it.

What it is and what it is not

What it is

A verifiable self-declaration: you declare the scope, we sign it, and anyone can cryptographically check that the declaration is yours, is current and has not been modified.
A W3C Verifiable Credentials 2.0 credential in vc+jwt format, signed with ES256, verifiable with any standard library without going through our servers.
A non-qualified trust service, provided by Cosmolabs OÜ, with a public and versioned trust framework.
A signed stop order towards your own systems, with an auditable record of every attempt.

What it is not

It is not an accreditation by an independent third party, nor a conformity review, nor an opinion on whether your agent is safe.
It does not attest that what is declared is true, nor that the agent behaves as stated: it attests who declared it, when, and that it has not changed since.
It is not a qualified trust service nor an electronic seal within the meaning of Regulation (EU) No 910/2014.
It does not make you comply with any rule. The obligation lies with the provider or deployer of the AI system, and it stays there.

CORTEX Seal is designed to provide a verifiable answer to the transparency obligation in Article 50 of Regulation (EU) 2024/1689: any third party can check, at a public URL, what the agent’s operator declared about its identity, its function and its limits. The legal obligation to inform rests with the provider or deployer of the AI system; CORTEX Seal does not replace it.

CORTEX Seal is a non-qualified trust service. It is not a qualified trust service nor an electronic seal within the meaning of Regulation (EU) No 910/2014, and it does not attest to the agent’s actual behaviour.

We call it, by analogy, the agent’s digital ID. CORTEX Seal is not an official document and has no connection with any national identity document or public authority.

On the roadmap: anchoring issuance and revocation in a distributed ledger. It is not deployed, it is part of no plan, and it will not be sold until it answers in production.

For agencies

Your clients’ agents, kept apart, with your brand on the seal.

The Agency plan is built for people who ship agents for other companies and have to show them to somebody with no account anywhere. What changes is not the volume: it is whose seal it looks like, and who can read it.

One workspace per end client: Each client lives in its own workspace, with its agents, its keys and its audit log. Nobody sees anybody else’s, and closing a project means revoking a workspace rather than going through a spreadsheet.
Your brand on the seal and on the page: The seal carries your agency’s logo and colour, and the verification page carries your identity and a link to your site. The page always states prominently that the issuer is Cosmolabs OÜ: presenting it otherwise would mislead whoever is verifying, which is exactly what the product exists to prevent.
A verifier on your own subdomain: As an add-on, verify.youragency.com. The issuer remains did:web:digitalcortex.tech and the page says so: the subdomain is presentation only.
Closing a project, in one click and with paperwork: When an engagement ends, you revoke that client’s agents and hand over the auditable record of the shutdown: what was revoked, when, which resources were cut, with what response code and how fast.
The manifest as a deliverable: The conversation about what the agent may and may not do is, in practice, part of the project scope. The manifest puts it in writing and signs it, and it can go into the proposal before signature.

See the Agency plan

Pricing

Public pricing. No sales call.

You pay per active agent in production, per month. Issuing, 90-day rotation, revocation and verification are unlimited within your plan: we charge for how many agents you have live, not how much you use them.

Currency

All prices are shown excluding VAT. At checkout, tax is calculated according to your country; for consumers the VAT of the country of residence applies. Cosmolabs OÜ is an Estonian company, registry code 17106582.

Sandbox

Free forever · no card

Anyone who wants to try the API and integrate before deciding.

€0

Unlimited sandbox agents
Unlimited issuance, rotation and revocation
A working public verifier
Dry runs of the stop button, with nothing executed for real
Seal stamped SANDBOX — NOT VALID

Start free Read the docs

Builder

Solo builder or product team with a handful of agents in production.

€41/month

Billed €490 yearly

5 agents in production
Unlimited issuance, 90-day rotation and revocation
Stop button over an HMAC-signed webhook
Audit log export in JSON and CSV
90 days of retention · 60 requests per minute

Start with Builder 30-day money back

Agency

Enterprise

From · annual only

Platform or consultancy that needs to issue under its own domain and its own DID.

From€1,490/month

Annual contracts only

Negotiated agent volume, with no cap in the catalogue
Delegated issuer under your own domain (did:web:your-domain)
Verifier on your own subdomain
730 days of retention · 600 requests per minute · unlimited seats
Negotiated data processing agreement and a contractual service level agreement

Talk to us See what’s included

Public pricing, with no sales call on Builder or Agency.

30-day money back, no questions and no retention call.

Cancel from the dashboard in two clicks. No email to anyone.

See the full comparison table, row by row

What will it cost you?

Agents in production

0200

Are they agents for different clients?

Only production agents with a valid credential count. Sandbox is unlimited and free. If an agent stops having a valid credential, it stops counting that same day.

Recommended plan: Builder
Annual price: €1,280
Per agent per month: €8.89
Add-ons needed: 1 × Block of +10 agents in production
You save paying annually: €256

At this volume the numbers work both ways. We show you both and tell you which is cheaper. Agency: €2,490.

Start with Builder See the full table

The rules, written down

0 agentsSandbox
1–5, a single clientBuilder
1–5, several clientsAgency
6–15, a single clientBuilder with +10 blocks, or Agency: both are shown
16–30Agency
31–80Agency with +10 blocks
81–150Agency with +50 blocks, plus a note recommending Enterprise
More than 150Enterprise

Add-ons

You add them from the dashboard and they prorate like the plan. Only what you can use today appears here.

Block of +10 agents in production

€790 / year

€7.90 per agent per month.

Builder · Agency

Block of +50 agents in production

€2,990 / year

€5.98 per agent per month.

Agency

Verifier on your own subdomain

€990 / year

The issuer remains did:web:digitalcortex.tech and the page says so prominently. The subdomain is presentation only.

Agency

Log retention: +12 months

€490 / year

On top of the retention your plan already includes.

Builder · Agency

Human review of the manifest

€390 one-off

€1,490 · pack of 5

A person reviews the permitted scope, the forbidden actions and the proposed autonomy level, and returns a report with recommendations. It is a professional service: it is not an audit and it attests to nothing.

All plans, Sandbox included

Assisted onboarding and migration of 5 agents

€490 one-off

Included at no cost with annual Agency and in the Design Partner programme.

All plans

30-day guarantee

Thirty days. If it is not for you, we refund you — no questions and no retention call.

It applies to Builder and Agency, monthly and annual, on each workspace’s first subscription.
On annual plans we refund in full within 30 days and, after that, the unused months pro rata during the first three months.
One email to support is enough. There is no cancellation form and no exit survey.
Production credentials already issued are not revoked: they expire naturally after 90 days, so you get a full quarter to land.

Professional services already delivered are not refunded: a human manifest review and assisted onboarding are somebody’s hours, already worked.

Design Partner programme

20 of 20 places left

Twenty agencies, €149 a month, locked for as long as the subscription stays active.

The full Agency plan at €149/month or €1,490/year, with assisted onboarding, migration of up to five agents, a verifier on your own subdomain for six months and one human manifest review, all at no extra cost.

What we ask in return

Your logo on the customers page.
One publishable case at 90 days: which agents, what scope they declared, what their end client said.
A thirty-minute call once a month for six months.
Permission to quote you with name and role.

We do not ask for exclusivity, cold referrals to your client base, or any volume commitment.

Twenty places is the honest number: twenty half-hour calls a month is ten hours, half the time available. The programme closes on day 90 and the founder price does not come back.

Apply for a place

What exists today, without decoration

We do not yet have paying customers, recognisable logos or installed-base figures. We say so here because the alternative would be to invent them. This is what you can check for yourself right now.

Argos, live: Credential CRD-2026-0001, level A4, four forbidden actions published. Verifiable with no account.
digitalcortex.tech/verify/argos
Documentation before sign-up: The technical specification, the manifest schema and the trust framework are published before anyone registers.
API documentation
Press release: Distributed in September 2026 through a newswire agency. We do not count outlets that merely syndicated the wire.
nextai.pro/en/digital-id-for-ai-agents
Technical articles, ES and EN: A content cluster on verifiable agentic identity, with its English version.
nextai.pro/dni-digital-agentes-ia
Who answers for it: Cosmolabs OÜ, Estonia, registry code 17106582. In a trust product, showing who you are is part of the product.
Estonian business register

Frequently asked questions

Pricing questions

The first agent takes about eight minutes.

No card, no sales call and no form asking how many employees you have. You create the workspace, publish a manifest, issue in sandbox and open your own verifier.

Start in sandbox Read the documentation

A verifiable ID for your AI agents. With a stop button.

We won’t tell you. Look at it.

Manifest

Declared scope, not just identity

A public page your client actually understands

Revocation does something

Six steps and one red button.

Workspace and key

Manifest

Credential

Verifier and seal

Linked resources

Revoke, with evidence

What we promise and what we do not

What it is and what it is not

Your clients’ agents, kept apart, with your brand on the seal.

Public pricing. No sales call.

Sandbox

Builder

Agency

Enterprise

What will it cost you?

Add-ons

Block of +10 agents in production

Block of +50 agents in production

Verifier on your own subdomain

Log retention: +12 months

Human review of the manifest

Assisted onboarding and migration of 5 agents

Twenty agencies, €149 a month, locked for as long as the subscription stays active.

What exists today, without decoration

Frequently asked questions

There are free agent registries. Why would I pay for this?

Is this an accreditation? Does it help me with the AI Act?

I already pay for an identity provider. Does this overlap?

I am not giving you my model-provider keys.

What if you revoke by mistake and take down my production on a Friday?

A large cloud will do this for free in six months.

You are a very small company. What if you disappear?

What actually makes you different?

Pricing questions

What exactly counts as an "active agent"?

Do you charge per credential issued or per verification?

What happens if I exceed my agent limit?

Can I change plan mid-cycle?

What happens to my credentials if I stop paying?

How much do I save paying annually?

How does the 30-day guarantee work?

Do prices include VAT?

Why is there no free plan with production agents?

What about anchoring in a distributed ledger?

The first agent takes about eight minutes.