Zero Trackers, 27 Languages: Gewerkton’s Egress-Free Privacy Architecture

Most multilingual product sites quietly become data-routing systems. A page loads, and the visitor’s browser begins talking to analytics platforms, advertising networks, consent tools, embedded media services and assorted third parties. The visible website may belong to one company, but the traffic tells a more complicated story.

Gewerkton has taken a markedly different route. Its marketing site supports 27 languages, contains zero trackers, displays no cookie banner and uses a fully egress-free architecture. It is a useful privacy-engineering case study precisely because the result is not presented as a decorative privacy feature. It is an architectural decision: the site does not need to send visitor data elsewhere to perform its job.

That approach also reflects the product behind the website. Gewerkton is a voice-first construction documentation and defect management platform for global markets. It was born in the German market, where it has its deepest commercial integration through GAEB, REB, XRechnung and DATEV, but its language coverage and regional technology choices are designed for international projects.

The platform is in beta now, with a public beta planned for fall 2026. That status matters. The architecture is ambitious, but it should be examined as an evolving beta platform rather than treated as a finished deployment with years of production history.

A website that does not phone home

“Egress” is the traffic leaving a system for an external destination. On a conventional marketing site, that traffic may be created by analytics scripts, tracking pixels or embedded services. Each connection can disclose something, even when the page itself contains no form asking for personal details.

A fully egress-free site changes the default relationship. The visitor requests the site, and the site serves its content without turning the visit into a chain of third-party connections. Gewerkton combines that architecture with zero trackers and no cookie banner across a 27-language site.

The absence of a banner is significant because banners have become a visual shorthand for privacy while often sitting on top of complex tracking arrangements. A banner can describe data flows, ask visitors to approve them or provide controls around them. It does not eliminate the underlying flows. Gewerkton’s architecture starts one layer deeper: remove the tracker and the external connection instead of making the visitor negotiate with it.

This is less theatrical than a large privacy dashboard, but more structurally interesting. The privacy property follows from what the site does not connect to. It is not dependent on whether a visitor finds the correct toggle, understands several categories or revisits a preference panel.

Twenty-seven languages make the decision more notable. International sites can accumulate technical dependencies as they expand, particularly when content, media and measurement are handled by separate systems. Gewerkton has paired wide language coverage with an egress-free design rather than accepting external traffic as the unavoidable cost of localisation.

Privacy without an empty website

Removing trackers does not require reducing a site to a placeholder. Gewerkton has a media bank of more than 51 self-produced clips and posters. The relevant architectural point is that a content-rich presentation and an egress-free approach are not mutually exclusive.

Self-produced material also gives the company direct control over what it publishes. The media bank can serve the explanation of the product without making third-party tracking part of the visitor’s viewing experience. In this model, marketing content remains content rather than becoming an excuse to add another external data relationship.

The site’s 27 languages are equally central to the case. Language is not merely a presentation setting for this platform; it is part of the operational problem Gewerkton is addressing. Cross-border construction teams may work across the EU, the US and APAC on the same project. The platform is intended to let each participant work in their own language while keeping the evidence original unambiguous.

That makes the marketing architecture consistent with the product story. A global audience can read about the platform without first being routed through a collection of tracking services, while project teams are offered choices about where their operational data resides and which regional AI providers they use.

Evidence changes the privacy calculation

Gewerkton’s marketing line is: “On site, what counts is what’s proven.” That sentence has architectural consequences. A platform whose purpose is to turn events on a construction site into evidence cannot treat custody as an administrative detail added after capture.

Construction records may connect a spoken instruction, a photograph, a deadline, a report, a plan or a handover. Their value depends on retaining a clear relationship between what happened, what was recorded and what the project team can retrieve. The more closely a system works with that evidence, the more important it becomes for its operators to know where the associated data is held and which providers participate in processing it.

This is not an argument for fear. It is an argument for architectural legibility. Project organisations should not have to discover their data path by reading a chain of vendor terms after deployment. Residency and provider selection belong near the beginning of the system design because they determine who operates the relevant infrastructure and where those operations take place.

Gewerkton’s data-residency choice is direct: use an EU cloud or run the platform on your own infrastructure. Those are materially different operating models, but the point is that the customer chooses between them. Data custody is therefore represented as a platform decision rather than a fixed consequence of buying the software.

Residency and AI region are separate controls

Data residency and AI-provider selection are related, but they should not be collapsed into one vague promise. Gewerkton offers the EU cloud or deployment on the customer’s own infrastructure for residency. Separately, its BYO-AI model supports 13 AI providers, lets customers bring their own keys and makes provider regions selectable across the EU, the US and Asia, including mainland China.

That distinction is useful. The location of the platform’s data is one architectural question. The selection of an AI provider and its region is another. Exposing both choices gives project operators a clearer way to shape the system around their geographic and organisational requirements.

Bringing your own keys also changes the relationship with the AI layer. Instead of making one bundled provider an inseparable part of the platform, Gewerkton supports a range of providers and explicitly avoids vendor lock-in. A project involving European, American and Asian participants can select a regional provider rather than being forced through a single global default.

Projects in Asia illustrate why the option matters. Chinese, Korean and Vietnamese crews may need multilingual operation from capture through to report, alongside data residency by choice. Regional AI-provider selection acknowledges that “international” is not a single technical region and that mainland China cannot simply be treated as an interchangeable extension of another market.

The architectural value lies in keeping the decisions visible. Residency, provider, region and keys are not synonyms. They represent different forms of control, and a credible data stance should keep those boundaries understandable.

Cloud as the coordination layer

The product line carrying this architecture story is Gewerkton Cloud. It handles operations and model/data coordination between Gewerkton Field, Gewerkton Studio and third parties. That makes Cloud the natural place for custody and coordination to meet.

Coordination platforms sit between people, project records and external participants. They do not merely display information; they establish how information moves between operational contexts. When Field captures evidence and Studio works with plans and models, Cloud connects those activities with each other and with third parties.

That role makes data choices part of the operating model. If Cloud coordinates records between systems, the organisation needs a deliberate answer to where that coordination environment runs. Gewerkton’s answer is the choice between an EU cloud and the customer’s own infrastructure.

The option to use your own infrastructure is particularly relevant for organisations that want the coordination layer inside an environment they operate. The EU-cloud option provides the other stated route. Gewerkton does not reduce all customers to a single hosting arrangement, just as its BYO-AI design does not reduce them to a single AI vendor.

This flexibility is not the same as having no external relationships. Cloud is expressly designed to coordinate with third parties, and the product supports multiple AI providers. The important architectural point is controlled participation: external systems are part of an intentional operating design, while the public marketing site remains fully egress-free.

From site capture to shared models

Gewerkton Field is the voice-first construction-site app. It handles dictation to evidence, defects, daywork reports, takt and portal functions. That places it close to the moment when a site event becomes a record.

Gewerkton Studio is the browser workspace for plans and models. Where no model exists, the site team can create one in the browser. Cloud then provides the operations and model/data coordination connecting those environments and third parties.

Together, the three lines describe a flow without turning the platform into an indistinct bundle. Field captures site activity. Studio provides the browser environment for plans and models. Cloud coordinates the operational and data relationships between them.

That separation also clarifies why Cloud carries the privacy architecture story. Capture may begin in Field and model work may happen in Studio, but coordination determines how those records participate in the wider project. The data-residency decision applies at the level where shared operations are organised.

Architecture across difficult project environments

The platform’s stated deployment fields show why a single central assumption would be limiting. Wind farms and renewable-energy projects can involve distributed sites, rotating crews, field acceptance and offline capture in dead zones. Data centres and industrial plants may have many trades working in parallel under tight deadlines, with meeting decisions converted into trade-sorted task lists.

Housing and building construction brings another pattern: defects recorded with photographs and deadlines, dictated daywork reports and signatures on the device at handover. Infrastructure and tunnel projects can run for long periods, accumulate many change orders and require instructions to be supported by original audio.

These are different environments, but they share a need to preserve the relationship between field activity and its evidence. A spoken instruction is not valuable merely because speech recognition produced text. The original audio can remain important. A defect is more useful when it stays connected to its photograph and deadline. A meeting decision needs to survive its conversion into work organised by trade.

Cross-border projects add language and regional infrastructure to the problem. Gewerkton supports teams across the EU, the US and APAC working on the same project in their own languages, while the evidence original remains unambiguous. Its 27 content languages and provider-region choices address that operational reality from different sides.

The result is an architecture built around explicit boundaries: local capture where connectivity may fail, multilingual operation across crews, coordination through Cloud, residency in an EU cloud or on customer-operated infrastructure, and AI selection across 13 providers in multiple regions.

Built by agents, tested with counterexamples

Gewerkton is being built by a solo founder directing a fleet of coding agents using Codex and Claude. In one night, that fleet shipped 21 software packages. The packages were verified with negative controls and mutation tests.

That development model is relevant to the architecture story because machine-assisted speed can magnify both output and mistakes. The notable fact is not simply that 21 packages were produced overnight. It is that verification included negative controls and mutation testing, techniques intended to challenge whether tests detect incorrect behaviour rather than merely confirming an expected path.

This does not turn beta software into a finished platform, nor does it settle every question about future operation. It does show an effort to pair accelerated implementation with tests designed to fail meaningfully. For a platform concerned with evidence, that is the more interesting part of the story.

Privacy as a property of the system

Gewerkton’s marketing site offers a compact demonstration of a broader principle. Privacy becomes easier to understand when it is expressed through architecture. Zero trackers is clearer than a long list of tracking purposes. Fully egress-free is clearer than a page that silently contacts several outside services. No cookie banner is credible here because it accompanies those structural choices, not because the banner itself has been hidden.

The product extends the same preference for explicit choices into data operations. Customers can choose the EU cloud or their own infrastructure. They can bring their own AI keys and select among 13 providers across EU, US and Asian regions, including mainland China. Cloud coordinates the data relationships between Field, Studio and third parties instead of pretending those relationships do not exist.

There is still an important temporal qualifier: Gewerkton is in beta, and its public beta is planned for fall 2026. The architecture should therefore be watched as it moves towards broader availability.

Even at this stage, the design presents a coherent thesis. A site can communicate in 27 languages without tracking its visitors. A global platform can expose residency and provider-region decisions instead of burying them. And software whose job is to preserve what happened on site can treat custody as part of the evidence architecture itself.