Vecto — Privacy Policy

Version: 1.1 · Effective: 2026-09-04 · (1.0 was effective 2026-08-23)

We built Vecto to be private by design, and this document is written to prove that rather than to assert it. It describes what actually happens in the system, including the parts that are not flattering, because a privacy policy that promises what the software does not do is worse than no policy at all.

If you only read one thing, read "Who else sees your data" and "Where your data lives".

What this policy covers

Three things, and they work so differently that mixing them up would make the rest of this document misleading:

• The Vecto service — the hosted platform at vecto.build and app.vecto.build. Everything in this document is about that, unless a section says otherwise.
• Vecto Proof — the signing service at proof.vecto.build, where you and other people sign a document together. It holds different data, keeps it for longer, and deletes it differently, because a signature has to stay provable. Its section is "Vecto Proof — documents you sign with someone else".
• Vecto Reader — a program you install on your own computer. No server of ours is involved when you use it. Its section is "Vecto Reader — software on your own machine". If you arrived here from the Microsoft Store, that is the one you came for, and you can stop reading after it.

All three are operated by the same company, named below.

---

Who is responsible

ITbrouwerij BV, Breendonkstraat 342a, 2830 Willebroek, Belgium. VAT BE0508.516.956 · RPR Antwerp, Mechelen division — "we".

ITbrouwerij BV operates the Vecto service and Vecto Proof, and publishes Vecto Reader in the Microsoft Store. If responsibility for either ever transfers to another company, this section changes and the change is listed under "Changes to this policy".

Contact for anything in this document: [email protected].

We have a certified Data Protection Officer on the founding team, who assesses our processing internally. We have not appointed a formal external DPO.

Two different roles, and which one applies

• We are the controller for data about you as a user or visitor: your account, our website analytics, our marketing contacts, our security logs. This policy is our notice to you for that.
• We are a processor for the content you put into your workspace — your product data, your documents, your source code, and any personal data those happen to contain. There, you decide what goes in and why; we only handle it to run the service for you. If you need a written data processing agreement under Article 28 GDPR, ask and we will put one in place.

The distinction matters most in the section on AI engines below, because that is where your choice as controller determines who else sees your content.

---

What we collect, and why

1. Your account

2. Your workspace content

Everything you or the AI create in Vecto: ideas, stories, epics, documents, decisions, plans, files you upload, source code you import or generate, chat conversations with the AI, and the graph that links it all together.

We process this only to provide the service. We do not mine it, sell it, or use it to train AI models. A free "community" tier in which insight into your data would be exchanged for free access is designed but not built; if it ever ships it will require your explicit, separate, revocable consent, and a paid alternative that shares nothing.

If your content contains other people's personal data, you decide that and you are the controller for it.

3. Visitors to our website (Vecto Pulse)

We measure how people find and use our websites with our own analytics — not Google Analytics, not any third-party product. Nothing about your visit leaves our systems.

This covers vecto.build and app.vecto.build, and the public pages of the sister products: reader.vecto.build, proof.vecto.build and sign.vecto.build. Each is recorded under its own name, so a visit to one is never counted as a visit to another. It covers the web pages only — never the programs. Vecto Reader installed on your machine sends nothing at all; see its own section below.

We set no cookies and store nothing on your device. No cookie, no localStorage, no tracking pixel, no fingerprinting.

For each page view we record: the page path (query strings and fragments stripped), the referring site's hostname (not the full URL), UTM campaign tags if present, your country as a two-letter code, and a coarse device class (desktop / mobile / tablet / bot).

To count unique visitors without knowing who they are, our server computes a hash of your IP address + your browser's user-agent + the site host, salted with a secret that rotates every day. The result is stored; the IP address and the full user-agent are not — they exist only in memory for the instant it takes to compute the hash and read the country. Because the salt changes daily, the same visitor produces a different value tomorrow, so we cannot follow anyone across days.

We also record whether a visit to the marketing site preceded a signup, so we can see how well the site works. That link carries no user ID and is derived the same cookieless way.

Legal basis: legitimate interest — understanding how our own product is used, in the least intrusive way we could find. You can object; see Your choices below, and please do if you want to, because we take that seriously enough to have built it into the collector itself.

How long we keep it. Raw events are kept for as long as we run the platform. They carry no IP address and no identifier that survives the day, so after that day there is nothing in them that points back to you — which is why keeping them costs you nothing and why we do not claim a deletion date we would not be able to prove.

4. If you contact us or subscribe

• Newsletter — your email address and language. We use double opt-in: nothing happens until you click the link in the confirmation email. Every email has an unsubscribe link.
• Demo request — your name, email, company, phone if you give it, and your message.

Both are sent by email to our founders, and both are added to our sales CRM, which is synchronised with a Slack workspace we control. Legal basis: consent for the newsletter; for a demo request, taking steps at your request prior to a contract.

5. Security and operational records

Vecto keeps an audit trail — it is a core feature, not incidental. It records who did what, when, in which workspace: logins, permission changes, AI dispatches, builds, spend, secret access, document changes.

Some of these records contain your IP address and browser user-agent. Specifically:

• Successful login and logout — IP and user-agent are stored, and are readable by administrators of your platform. Retained 30 days.
• Failed logins, throttled login attempts, and password-reset requests — IP, user-agent and the email address that was typed (even when no such account exists) are stored in a separate security log.
• Giving a connected service permission to see your details — the IP address you gave that permission from is stored with the permission itself, so that a later dispute about who agreed to what has an answer. Withdrawing the permission does not erase it: the record that permission once existed is part of the trail of what a service was allowed to do, and it is kept with the withdrawal marked on it. It goes when your account goes.

Read this one carefully: the security log is never deleted. It is retained indefinitely, by design, because an audit trail that can be erased is not an audit trail. We consider that necessary for security and accountability, and we would rather tell you than let you assume otherwise.

Reading the contents of security-sensitive audit records requires a separate auditor permission that is granted deliberately and is not attached to being an administrator. Without it, even a platform administrator sees only the headline of such an event, not its payload. Automated access tokens never carry it.

Legal basis: legitimate interest (security, fraud prevention, accountability) and legal obligation.

6. Error reports

When something breaks, we record the error so we can fix it: the error message and stack trace, the page URL, the route, the browser user-agent, your user ID, your workspace, and a limited context object.

Before anything is stored, it passes through a scrubber that removes credentials — authorisation headers, cookies, API keys, access tokens, JWTs, private keys, and anything matching a known secret pattern — and that keeps only an allow-list of HTTP headers. Query-string values that look like secrets are redacted while the parameter names are kept.

We do not store your IP address in error reports. Error events are deleted after 30 days; the aggregated error signature is kept for 365 days, or indefinitely if we promoted it into a tracked issue.

Legal basis: legitimate interest (keeping the service working).

---

Who else sees your data

This is the section most policies make vague. Here is the actual list.

Always in the path, whatever you configure

We are not happy about the last two and intend to self-host those assets. They are listed because they are real: loading a font from a CDN discloses your IP to that CDN, and pretending otherwise would be dishonest. Until we fix it, this is what happens.

Email

AI engines — this one depends on your configuration

Vecto can run on a local model on our own hardware, or on a hosted model from Google (Gemini), OpenAI, Anthropic, xAI or Mistral. Which one runs is a setting, per workspace and per capability slot, and it determines whether your content leaves our infrastructure at all.

When a hosted engine is configured, your content is sent to that provider. That is not limited to prompts. When the AI works on an existing codebase, it reads your actual files and sends their contents to the model as part of the conversation — that is how it can edit real code rather than guess. Documents, product data and chat messages go the same way.

There are three arrangements:

1. Your own key. You supply your API key; you have a direct relationship with the provider and their terms and privacy policy apply to you.
2. Our key, metered. We call the provider on your behalf and record the usage and cost.
3. The local model. Inference runs on hardware we operate. Your content does not go to any AI provider at all.

A first build of brand-new code often runs as a scripted replay with no AI call at all. But be aware of one honest caveat: if a local model fails, the system can fall back to a hosted engine. That is a known open item, tracked internally, precisely because "it stayed local" should never be something you have to hope for. Which engine actually ran each step is recorded and visible to you.

Only if you connect them

Who does not receive anything

• No payment processor. Billing is manual; no payment provider is integrated and no card data ever reaches us.
• No third-party analytics. No Google Analytics, PostHog, Mixpanel, Segment, Plausible or Matomo.
• No third-party error tracking. No Sentry, Bugsnag, Datadog or New Relic. Errors go to our own database.
• No external login provider. No Google/Microsoft SSO — authentication is entirely ours.
• No advertising network, no data broker, ever.

---

International transfers

The platform runs on our own infrastructure in Europe. Three things send data outside the EU:

1. Cloudflare operates a global edge network, so traffic may be handled outside the EEA.
2. Hosted AI engines — Google, OpenAI, Anthropic, xAI — are United States companies, and content sent to them is processed in the US or wherever they operate.
3. Resend, a United States company, sends the email that originates from the public website — the newsletter confirmation and the demo-request confirmation. That is your email address and the text of that one message; it is not the channel your in-product email uses.

This is the practical consequence of the engine choice above, and it is worth being blunt about it: if you configure a US-hosted AI engine, your product data and your source code are transferred to the United States. If that is unacceptable for your data, use the local model.

Where such transfers happen, they rely on the European Commission's Standard Contractual Clauses and, for the US providers concerned, the EU–US Data Privacy Framework where applicable. If you use your own provider key, the transfer is between you and that provider under their terms.

Google Fonts and unpkg receive your IP address as part of loading page assets, from wherever their CDN serves you.

---

Where your data lives, and how it is protected

We would rather describe our security accurately than impressively.

Hosting. Vecto runs on hardware we own and operate ourselves in Belgium — not on AWS, Azure or Google Cloud. It is not directly reachable from the internet; all traffic arrives through a Cloudflare Tunnel. The hardware for the local AI model sits on the same machine, which is why choosing that model keeps your content on our infrastructure.

What is protected, and how:

• Passwords are hashed with bcrypt (cost factor 12). We cannot read them. Repeated failed logins trigger both a per-account lock and a per-IP ban.
• Secrets and stored credentials — your API keys, connected-service credentials, and anything you keep in Vecto's secret store — are encrypted with AES-256-GCM before they touch the database. Secret values are never returned by the general read APIs; a server-side allow-list strips them out, and there is exactly one audited path to retrieve a value.
• In transit, traffic is encrypted with TLS terminating at Cloudflare's edge.
• Access is controlled by role and capability, checked on every request. Authentication is enforced centrally, with a deliberate list of public exceptions — sign-in, health checks, looking up an invitation by its token, the newsletter and demo forms, error telemetry, the analytics beacon, the endpoints a sister service uses to sign you in, and profile pictures. That last one is the only exception that serves a personal detail: a profile picture is reachable without signing in, at an address that cannot be guessed and that changes whenever the picture does. No route reaches workspace data without an authenticated session or a scoped access token.

What is not protected the way you might assume — stated plainly:

• The main database is not encrypted at rest. Only secrets and credentials are individually encrypted; the rest of the data sits on a normal disk. Enabling storage-level encryption is a committed, tracked item that has not shipped.
• The key that protects those secrets is held on the server itself, not in a hardware security module or a managed key-management service. Moving to envelope encryption with a managed key is designed but not built.
• Separation between customers is enforced by our application, not by the database. Every workspace's data lives in the same database, and each query filters on the workspace. There is no database-level row-level security acting as a second net. The code is written carefully and concentrates these checks in shared helpers, and an authenticated multi-identity penetration test of cross-tenant access was run in June 2026 — but the honest description is logical isolation, not physical or database-enforced isolation.

We keep a fuller, technical account of our security posture — including the specific gaps we are working through and the evidence behind each control — in our internal compliance documentation. We share it with customers who need it for their own risk assessment; ask us and we will walk you through it. We do not publish the implementation-level detail, because a public inventory of exactly where our defences are thinnest helps an attacker more than it helps you.

Backups are taken daily to encrypted storage under our own control, in Belgium. Data deleted from the live system stays in those backups until the snapshots holding it age out, so erasure is complete in the live system immediately and in the backups on that cycle.

Security testing. We assess our public attack surface ourselves, record what we find, and record the risks we have deliberately accepted rather than quietly leaving them out.

We are not ISO 27001 certified. We maintain internal compliance documentation modelled on ISO/IEC 27001:2022 — control objectives, evidence, and an honest gap register per pipeline — as preparation for a future certification. That is preparation, not certification, and we will not imply otherwise.

---

How long we keep things

Vecto Proof keeps things longer, and on purpose. Signing records have their own retention, with a floor of ten years — see "Vecto Proof — documents you sign with someone else". The table above is about the Vecto service.

---

Your rights

Under the GDPR you can ask us to give you a copy of your personal data, correct it, delete it, restrict or object to how we use it, and receive it in a portable format. You can withdraw consent at any time where consent is the basis.

How to exercise them: email [email protected]. We will respond within one month. We do not charge for this.

How this works in practice:

• Access and portability. Your workspace content is retrievable through our API and MCP server, documents are plain Markdown, and files can be downloaded. Ask us and we produce a complete extract for you.
• Deletion. You can do this yourself: Settings → Profile, where it asks for your password and for your email address typed out in full. Both are deliberate — erasure cannot be triggered by someone who merely found your screen unlocked. You can also write to us and we carry it out for you. Either way you are told when it is done, and what stays behind is listed below.
• In Vecto Proof, erasure has a limit that is written into the product. Your content goes; the proof that something happened stays, because it is also the other signer's evidence. The section on Proof sets out exactly what is destroyed and what is not.
• What survives deletion, and why. Security audit records, and audit records generally within their retention windows, are kept for security and accountability. Backup snapshots age out on their own schedule. Aggregated, non-identifying counts are kept.
• Marketing contacts. For newsletter subscribers, demo requests and CRM contacts, we have a working erasure function that scrubs name, email, phone and notes and leaves only an anonymous tombstone.
• Objecting to analytics. See below — this one you can do yourself, immediately.

Complaints. If you are unhappy with how we handle your data, please tell us first — we would like the chance to fix it. You also have the right to complain to the Belgian Data Protection Authority (Gegevensbeschermingsautoriteit / Autorité de protection des données), Drukpersstraat 35, 1000 Brussels — gegevensbeschermingsautoriteit.be, or to the supervisory authority where you live.

---

Your choices

Opting out of our website analytics

Because we never store anything on your device, we cannot offer you an opt-out cookie — that would be self-defeating. Instead our server honours the signal your browser already sends:

Turn on "Do Not Track" or "Global Privacy Control" in your browser and we stop measuring you. When a request arrives with DNT: 1 or Sec-GPC: 1, the event is discarded before anything is parsed, hashed or written. Nothing about that visit is recorded — not a page view, not a hashed visitor value, not a conversion.

Most browsers offer this under Settings → Privacy. Firefox, Brave and Safari have it built in; in Chrome and Edge, privacy extensions can send the Global Privacy Control signal.

If you would rather object in writing, email [email protected].

Choosing where your content is processed

The AI engine setting is yours. If you do not want your content going to a US provider, configure the local model, or supply your own key so the relationship is directly with a provider you have vetted.

Email

Every newsletter email has an unsubscribe link. Service emails — password resets, invitations, security notices — are part of the service and are not marketing.

---

Cookies and device storage

The public website sets no analytics or advertising cookies.

Vecto uses a small number of strictly functional items:

Inside the application the browser's local and session storage also hold transient interface state — for example a pending invitation token while you switch accounts. None of it is analytics.

None of these track you, none are shared, and all are necessary for the site or app to work as you asked it to. That is why there is no consent banner: there is nothing to consent to.

---

Vecto Proof — documents you sign with someone else

Vecto Proof is a separate service at proof.vecto.build. You put a document on the table, invite the people who must sign it, and each signature is recorded so that it can still be proven years later. That last requirement is why this section exists: a signature that can be erased is not a signature, and the consequences of that run through everything below.

There are no organisations in Proof. Everything belongs to a person.

How you get in

You sign in with your Vecto account, or by clicking a link we mail to your address. Anyone can receive an invitation to sign; you do not need an account of your own to sign something, and you cannot create an account yourself — accounts exist by invitation.

What Proof holds about you

• Your account — your address, your name if you gave one, and your sign-in sessions.
• The people you invite — their name and address, because that is who the invitation goes to and who the signature will be attributed to. If you type someone's address in, you are giving us their data; you decide that, and this policy is our notice to them as much as to you.
• The documents themselves, and every earlier version that anyone signed.
• The record of what happened — who was invited, who proposed a change, who declined and why, who signed and when. This is the evidence, and it is described below.

Your documents are encrypted, and here is the limit of that claim

Documents are stored encrypted. What we promise about it is exactly this and no more:

We cannot read your documents without it leaving an indelible trace. Not "we cannot read them" — the master key is held by us, so the honest claim is traceability, not impossibility.

If we ever open a document, that opening is written into your proof record, where you can see it. We would rather tell you where the boundary of the guarantee sits than let a stronger sentence do work it cannot carry.

When everyone has signed

Each signer who is not the owner is sent the finished, signed document by email. This is the one place where a signed document deliberately leaves our encrypted storage on our own initiative, and it is a choice rather than an oversight: the person who signed something is entitled to hold their own copy of it, without needing an account with us to keep it.

The certificate and the timestamp

To sign, a certificate is issued in your name by Vecto Sign, our own certificate service, and the signature is stamped by our timestamp authority so that it stays verifiable after the certificate expires.

What the certificate contains, and what that means:

• Your name and your country, as they were supplied. Neither was checked against an identity document. The certificate says so itself, in the level it is issued under.
• Your email address, and that one is evidence. At some point post from us demonstrably arrived there — a redeemed reset link, an accepted invitation, or a confirmed address change. That proof is recorded against the address it was given for, and it moves with the address when you change it through the proper route. It is not re-established at the moment you sign.
• An identifier of our own for the subject — not a national registration number.

A certificate that has been issued cannot be recalled or altered, because everything signed under it must remain verifiable. The issuing service keeps a record of what it issued and when, for as long as any signature made under it may still need checking.

Deleting a dossier, and what does not go

You can delete a dossier, and when you do, the content is destroyed: the document text, every signed version, the finished artifact with the signature block in it, and the contents of the earlier entries in the record — proposals, reasons for declining, addresses inside them.

What remains, and why it is not ours to remove:

• The chain of the record itself — the order of events, their hashes and their timestamps — so that a proof record fetched afterwards still verifies.
• Who signed and when. The signer's address is part of what was hashed; removing it would break the very thing it proves.
• The fingerprint of what was signed, and an entry naming what was destroyed — otherwise the record would say that something was destroyed without ever saying what, and the other party's own copy would have nothing to be laid against.
• The signature, its timestamp and the certificate chain. That is evidence about the document rather than the document.
• The invitation link for a signer. It is how they reach their own proof record, and taking it away would remove exactly what they were promised.

This is the honest shape of erasure in a signing product: your content goes, the proof that something happened stays. The counterparty's evidence is theirs, not yours to make unverifiable — and the same protection works the other way round when you are the one who signed.

How long

Each account has a retention period, which you can set, with a floor of ten years. It runs from the last thing that happened in a dossier, and only once every request in it is closed. The floor exists because documents people sign are the kind that get asked about years later.

---

Vecto Reader — software on your own machine

Vecto Reader is a program you install on your own computer. It opens markdown files and checks the signatures inside them. Everything in this section works differently from the rest of this policy, because no server of ours is involved. Nothing above about our server, our hosting, our sub-processors or our retention periods applies to it — not because we carved out an exception, but because there is no server in the picture at all.

The file you open

It is read from your disk, rendered and checked on your machine. It is not uploaded, not copied anywhere, and never sent to us. Opening a file makes no network request at all.

What the reader stores, and where

On your machine only. There is no account, so there is nothing to synchronise it to.

Images from outside the document

A document can point to an image on someone else's server. The reader does not load those unless you say so. If you do, that server learns your IP address and the moment you read the document. The choice is offered per document and is never made for you.

What we do not have

We have no way to know that you use Vecto Reader, which documents you open, when, or what they say. There is no account, no telemetry, no crash reporting and no update check. That is not a gap we intend to fill later; it is what the product is.

One distinction, because it would otherwise read as a contradiction. The download page at reader.vecto.build is a website, and we measure it the way we measure our others — page views, which package was downloaded, a country code. That tells us how many people fetched the program. It tells us nothing about anyone who runs it: the two are not connected, and the program itself never sends anything. If you download the Reader and never return to that page, we know a download happened and nothing further.

It also means the rights described under "Your rights" have nothing to act on here. There is no data of yours in our hands to give you a copy of, correct or erase — a request would come back empty, and that is the honest answer rather than a refusal.

What Microsoft sees

The reader is distributed through the Microsoft Store. Microsoft therefore knows that you installed it, and gives us aggregate figures — installs, ratings, and usage counts per market. We do not receive your identity, and none of it can be linked to a document you opened. Microsoft's own privacy statement governs that part, and we cannot switch it off: it is the price of distributing through a store instead of a download link.

---

Children

Vecto is a professional tool and is not directed at children. You must be at least 16 to have an account. We do not knowingly collect data from children; if you believe we have, tell us and we will delete it.

---

Automated decision-making

Vecto generates software automatically, and it makes autonomous technical decisions when you allow it to. It does not make automated decisions that produce legal effects concerning you or similarly significantly affect you, and it does not profile individuals. Our website analytics are never used to make decisions about a person.

---

Changes to this policy

We update this policy when the platform changes. Material changes are notified to account holders before they take effect, and every version is kept, so what changed and when is traceable.

Version 1.1 (2026-09-04) — three additions, no reduction of anyone's rights. The analytics section now names every website it covers, including the public pages of Vecto Reader, Proof and Sign, and states that it covers those pages and never the programs. The Vecto Reader section says the same thing from the other side, so that "we cannot know that you use Vecto Reader" is not read as covering the download page. And the security-records section names a third kind of record that holds an IP address: the one written when you give a connected service permission.

---

Contact

ITbrouwerij BV Breendonkstraat 342a, 2830 Willebroek, Belgium VAT BE0508.516.956 · RPR Antwerp, Mechelen division Email: [email protected] Web: https://vecto.build

The same company publishes itbrouwerij.be and its own privacy statement. That one covers the consultancy; this one covers Vecto and Vecto Reader. Where they describe the same company, they say the same thing — if you ever find them disagreeing, this document is the one about Vecto and the discrepancy is a mistake worth reporting.

---

Every factual claim in this document was checked against how the software actually behaves, not against how we would like it to behave. If you find a sentence here that the product contradicts, tell us — that is a defect in the product or in this page, and either one is ours to fix.