Sovereign AI,
delivered locally.
Agents you can trust.
Adopt useful AI without an engineer — a consultant you already trust installs and runs it, and your knowledge never leaves the building.
Animation: a customer question is grounded against the knowledge base and answered by the local AI core.
You hold the reins — governed projection and scoped, revocable access.
No black boxes — every answer cites its sources, so you can audit why.
Someone answers for it — Cloud Shield and a local consultant stand behind it.
Useful AI is locked away from the businesses that need it most
SMEs run on knowledge in people’s heads, scattered PDFs and shared drives. The AI options that actually work are out of reach — for three reasons.
Sovereignty lost
Turnkey cloud agents (Sierra, Decagon, Intercom Fin) send your proprietary and customer data to a foreign provider — outside your control.
The complexity wall
Self-hosting an open model with tuned retrieval and access control is an engineering project an SME cannot staff or maintain.
Compliance fear
DSGVO and the EU AI Act make adoption feel radioactive. Legal uncertainty is the #1 cited AI barrier — so the safest move has been to do nothing.
The only governed, sovereign-by-default AI assistant an SME can adopt without an engineer — because a trusted local consultant runs it for them.
The product is not retrieval. It’s governance.
Anyone can bolt a chatbot onto a pile of PDFs. The hard, valuable part is the boundary between private knowledge and a public-facing agent — and the loop that makes it improve.
- Governed projection. Knowledge is default-private and promoted to the public agent by review. The public surface can never leak the internal corpus by accident.
- One brain, many bases. A single local LLM serves many knowledge bases behind simple access controls — staff-facing and customer-facing, cleanly separated.
- The correction loop. Uncertainty escalates to a human; that answer becomes new knowledge. Good answers stop dying inside support tickets.
No path back: the public agent can never read into the internal corpus. The reverse data flow is blocked by construction — there is no setting that opens it.
The correction loop — the one path back: when the agent is unsure it escalates to a human, and that vetted answer is captured into the knowledge base. Knowledge only returns through a person, never through the agent.
Watch the correction loop compound
Every escalation becomes new knowledge, so the agent resolves more on its own each week — and never guesses to get there. This is the same conversation-analysis view your consultant watches in the console. Hover the chart to read a week.
- +28 pts
- resolved without a human, over eight weeks
- 0 guesses
- every answer is cited or escalated to a person
- Stays on your box
- the loop runs locally — nothing leaves the building
More resolved automatically — week over week
Resolved without a humanThe share of questions the agent answers without a human, as the correction loop learns from each escalation.
Illustrative — your numbers depend on your knowledge base.
A concierge, not a deflection bot
A deflection bot exists to make people go away. A concierge recognises a returning customer, remembers the relationship and tends it over time — and it can do that because identity, history and follow-up all live on your own box, never in a foreign cloud.
Passwordless customer accounts
A returning customer signs in with a magic link or one-time code — no password to forget. The account is scoped to your core alone, so the agent recognises them and carries their history without ever widening what it can see.
Sovereign on-core CRM
One contact record per person, with a merged timeline of chats, escalations and notes, plus tags and follow-ups — all stored on your own core. The relationship layer lives with you, never in a foreign cloud.
Identity projection
At answer time the agent injects only the authenticated customer’s own context — default-deny. It never reaches another customer’s data, and never the internal corpus. The trust boundary holds, per person.
Same governance spine — warmer, relationship-led.
Private AI, on your premises
The AI core — a local model plus your knowledge — runs inside your building and is used locally by your team. Cloud Shield manages it remotely over a single outbound connector tunnel — no inbound ports are ever opened. Only governance, compliance and signed updates flow; your data never leaves.
Diagram: the sealed AI core (appliance plus WegentyOS) — a local model and your knowledge — runs on your own premises and is queried locally by your team. The core dials an outbound, mutually-authenticated connector tunnel to Cloud Shield, the zero-trust public edge; no inbound ports are opened and no data leaves the building.
One box. Independent choices.
The hardware is the constant — a pre-imaged, sealed AI core inside your own network, and the one trust boundary that never changes. Everything else is an independent choice you stack on top: who sets it up, how (and whether) it is reachable from outside, and who tailors it.
The appliance
A pre-imaged, sealed AI core that plugs into your network. It runs WegentyOS — no inbound ports, signed updates — and is the trust boundary that never moves. One box, owned by you. Everything below is chosen on top of it.
Who stands it up
Wegenty, a local consultant, or do it yourself. Most owners let a consultant they already trust plug it in — installed in an afternoon.
Who can reach it
Pure local is always the baseline. Public exposure is additive and optional: a DIY public API and/or Cloud Shield — combinable on the same core, not an either/or. Stay fully local, or open up under governance.
Who tailors it
Do it yourself, have a consultant do it, or commission Wegenty as a premium option. Tune knowledge, agents and integrations to fit how you actually work.
Sovereignty is the floor, not an upsell.
Buy your AI core today
A pre-imaged appliance from ~€490, installed in an afternoon by a consultant you trust. Your AI runs on a box you own — no cloud bill, nothing leaves the building. Every appliance runs WegentyOS — a hardened, signed Linux image that seals the box and ships immutable updates. Appliance × WegentyOS = your AI core.
Built so a well-funded competitor can’t just copy it
Four reinforcing moats — distribution, compliance, sovereignty, and governance architecture.
Run by people, not ads
A trusted local consultant installs and runs it for you, and earns on a relationship they already own. Cloud incumbents are structurally self-serve — they cannot match a human trust network.
Compliance, carried for you
Cloud Shield is the closed compliance umbrella — live regulatory posture, audit trail and certification under DSGVO and the EU AI Act. The part no solo operator wants to build, run, or self-insure.
Sovereign + open
A local LLM, your knowledge under your control, an open-source core. Cloud-bound incumbents cannot copy this without cannibalising their own architecture and margins.
Governance, by design
Internal superset, public projection, never the reverse — plus a correction loop that compounds with use. Easy to describe, hard to operationalise safely for non-technical owners.
One platform, four people — and only one of them is the buyer
The consultant sells and runs it, the owner buys and owns the data, the developer validates the openness.
Recurring revenue on the trust you’ve already earned
Sell a sovereign, compliant AI assistant you didn’t have to build. Install it in an afternoon, keep the customer relationship, and earn a recurring share — the platform carries the liability.
Become a partnerYour knowledge, working for you — without leaving the building
Local services, retail, trades, hospitality — answer customers and staff with an assistant grounded in your own knowledge. Someone you trust handles it end-to-end. You stay demonstrably compliant, and never touch the plumbing.
Learn moreRecognised, remembered — a concierge, not a queue
Returning customers are recognised and their history carried, so answers feel personal — grounded, cited, in seconds. When the agent is unsure, it hands off to a person who already knows their account. No dead ends, no hold music.
Try the live demoOpen core, sovereign by default
Read the code, run the stack, probe the trust boundary. Everything is open source except the compliance layer — fork everything but the liability.
Explore the sourceThe EU AI Act is fully applicable on 2 August 2026
GPAI obligations are already live. For most SMEs the deadline reads as risk. Wegenty turns it into a decision you can make confidently — because Cloud Shield carries the live regulatory posture, audit trail and certification, and a local advisor stands behind it.
GPAI obligations have been live since August 2025; the EU AI Act is fully applicable on 2 August 2026; some high-risk (Annex III) duties phase in through 2027.
An honest free path, and a workhorse tier
Names, limits and pricing are illustrative at v0.1 — set with our first consultants.
Community
The full open-source core. No Cloud Shield, community support. The honest free path.
Get the sourceProfessional
Cloud Shield, one public agent, multiple knowledge bases, consultant onboarding, standard SLA.
Book a demoEnterprise
High availability, SSO, multiple agents, certification support, priority response.
Contact salesSee the full breakdown on the pricing page.
See your own knowledge answer a customer — sovereignly.
Try the interactive demo, or book a 20-minute walkthrough. No data leaves your building, ever.