Sovereign AI · Founded 2026

Sovereign AI, Software and AI solutions for your business.

There is growing interest in sovereign AI and ethics. I've started looking into ethics because at the moment I see AI causing more harm through job displacements than good, and it shouldn't be like that. The potential is immense. People and knowledge are essential in the age of AI. We provide the solution and act first to protect our people and country.

Owned by people, not rented from a few.

Want to know more, contact our team

What becomes possible

Autonomous AI, put to work.

Terranimus is among the early providers building autonomous AI for business. This is Elara — the visual TerranSoul takes on your screen. She is AI staff that plans a defined job, acts through the systems you already use, reports what happened, and brings a person in for approvals and exceptions.

Elara, a 3D companion

Ask her something
  • Answers in a chat window on your own site, and acts on it — not just talks about it.
  • Installed on your own computer, not rented from someone else’s cloud — the same TerranSoul stack, running local models, or any cloud provider you choose to connect, like OpenAI, Claude or Kimi.
  • You set the rules once — automate a task, teach your team how something works, or demonstrate it to a customer — and she keeps doing it without being asked twice.
  • Runs your internal workflows and integrations, and tells you what happened rather than making you go and check.
  • Shares what she knows over MCP, so tools like Claude Code and Codex can read it — and she can act through them directly, the same way she acts through your own systems.
  • Everything she remembers stays on your machine: your data, your infrastructure, your integrations.

No per-seat subscription — you already own the model, the memory and the machine it runs on.

What we built once, and can add to yours in days.

A booking system we built end to end — availability, promotions, calendar sync, a customer portal, an admin dashboard, and an assistant that can look things up but never books or cancels on its own. Build once, apply to all: the same pieces can be added to your existing app or site, typically within days.

  • Booking that checks itself

    Multiple services booked against live availability, with course and workshop registrations kept separate so a class signup never eats a service slot — plus a simple way for someone to register interest before committing to a booking.

  • Promotions that run themselves

    First-booking, birthday-window and review-reward discounts, each checked automatically against a customer’s own booking history, with a manual override for the cases a rule should not decide.

  • Calendar and meetings, kept in sync

    Calendar invites and meeting links generated on confirmation, updated automatically if a booking is redated, and withdrawn cleanly if it is cancelled.

  • A portal for the customer, a dashboard for the business

    Customers manage their own bookings, promotions and reviews. The business gets one dashboard for every booking, customer and promotion, including booking on a customer’s behalf.

  • Invoices and reminders, sent without being asked

    Confirmation emails carry an invoice and a calendar invite automatically, in whichever of seven languages the customer used to write in.

  • An assistant that looks things up but never acts alone

    Customers can ask about a booking or a promotion in chat. The assistant reads and explains, then hands cancellations and reschedules to a person to confirm — it does not book or cancel anything by itself.

Try it on our own memoryRather than describe it: the graph on that page is the memory our own assistant runs on, and you can talk to it and watch it change.

And where a job needs something built rather than embodied

  • Execution

    Tesseres

    The other product: an integration and execution engine. AI does the work; a person approves every change; every action is traceable.

    See it move
  • Bespoke

    AI staff for your business

    We also take on contract work — tailored systems and features, wired into what you already run, and built with current AI rather than whatever shipped last year. Prices and terms are on Services.

    Talk about a build

What we do

Most businesses do not have an AI problem. They have a process nobody has drawn.

Terranimus builds software, AI and automation onto the systems a business already runs. We start with the process, because AI bolted onto a broken one only makes the mess arrive faster. And we build it so the result is yours: the data, the infrastructure, the integrations and the workflows stay with you. That is what sovereign AI means once it stops being a slogan and turns into an invoice.

See how it works

Why it holds up

Four things that are true of everything we build.

  • Memory you can read

    The fix for an assistant that forgets between sessions and invents when the context runs long: memory kept as data rather than weights, so it can be inspected, corrected and exported. The guide on this page is built that way — ask it something this site does not publish and watch it decline.

  • A person releases it

    Nothing we build reaches your production systems without someone approving it. That is not a setting we switch on for nervous clients — it is the gate in the middle of the diagram, and the argument the company is built on.

  • Failures are not silent

    A call that fails takes a dead-letter path where it can be inspected and replayed back through the broker, rather than retrying quietly until something downstream gives way. You find out, and you re-run it.

  • Additive, never load-bearing

    Drop us and your software carries on doing exactly what it did before — it simply stops being AI-assisted. Nothing we add becomes structural to something that already worked without us.

What we build

One partner for the whole of it.

From a site that does a job through to multi-tenant platforms with isolation between customers and distribution across regions. Any scale you grow into.

  • AI staff

    An AI that works alongside your team on a defined job — reading, drafting, answering, filing — supported month to month rather than handed over and forgotten.

  • AI in your workflows

    AI placed inside a real process with a trigger, a branch, and a person who approves the consequential step, instead of a chat window someone has to remember to open.

  • Integrations

    Systems that were never introduced to each other, joined properly: a broker, a translator, a router, and a dead-letter path for the calls that fail.

  • Business process automation

    The quoting, invoicing, chasing and reporting that runs on somebody’s evening. Mapped first, automated second.

  • Custom AI agents

    Purpose-built for one job, grounded in your own material, and able to say "I do not know" — which is the difference between a tool and a liability.

  • Memory and retrieval

    The fix for an assistant that forgets between sessions and invents when the context runs long. Memory as data you can read, not weights you cannot.

  • Software, at any size

    From a site that does a job through to multi-tenant platforms with isolation between customers and distribution across regions — the same engineering, scaled as you scale.

All services and prices

How we work

Five steps. Nothing is payable until the fourth.

The consultation, the system design and a first working version — that one inside a working day — cost nothing.

  1. Consultation — free

    We map what you actually do, where the time goes, and what it costs you now. You get that map whether or not you hire us.

  2. System design — free

    The architecture, the integrations, the approval points and the failure paths, designed and written down before anyone asks you for money. It is yours to keep — and yours to hand to another builder if you would rather.

  3. A working demo — free, within a day

    One working day after the design is agreed, you get a demo app that runs on your own material: the slice that shows the design holds up. A day of build is the bound, and it is deliberate — it is not the finished system, it is enough to judge one. You will have watched the real thing move before you have paid anything, and only then do we quote.

  4. Fixed scope, fixed fee

    Only once you have watched it work do we put a scope and a number in writing, and only then does anything reach your real systems — with a person approving every change that does. No hourly drift, and no discovering the price at the end.

  5. Support and scale

    Ongoing support on a monthly subscription, and the same engineering as the system grows. What we built sits on top of what you already run: keep us and it gets better, drop us and your systems carry on without the AI layer.

See the prices

What it costs

Nobody will tell you the price, so here it is.

What an AI or software project should cost, and how long it should take, is not obvious to anyone — not a small business asking us for the first time, and not a large one running its own engineering team. Guesses for the same piece of work can be an order of magnitude apart, which is exactly why so much of it never starts. Our answer is not to guess better — it is to stop building everything from zero. The R&D for a system like TerranSoul or Tesseres happens once and gets applied to every client after that, rather than reinvented for each one, which is what actually lowers the cost, shortens the time and improves the quality: a system already made to work properly, not one being built for the first time on your budget.

  • From $100 to $10,000 and beyond — the bands are on the page, not behind a call.
  • Free before anything is payable: the consultation, the system design, and a demo app running on your own material inside one working day.
  • Tell us your budget and the advice is built around that number — including telling you when it is too small for what you are asking for.
See every price

For agencies and consultants

Build once, apply to all.

If you already run an agency, a studio or a consulting practice, we build the website, app, software or AI underneath your service at wholesale pricing, so the same build can go out to every client you bring us, not just one. You sell it under your own name. We work for your client, never with them directly.

  • Wholesale pricing, meant to be resold rather than a one-off project rate.
  • Built to your brief and your brand — your client never sees our name on it.
  • We never contact your client directly, before, during or after the build.
Talk about wholesale

What we build

  • AI Booking System

    Customers book a time online and an AI assistant answers their questions. A person handles changes and cancellations.

  • Landing Page

    One clear page that says what you do and gives visitors an easy way to contact you.

  • FrontStore

    An online store where your catalogue, enquiries and orders join your back office in one flow.

  • Marketplace

    Many sellers in one storefront, built as a personalised app and quoted per app.

Before you ask

The questions we hear most.

I do not know where AI fits my business.

Most of the time it does not fit yet, because the process underneath it is the problem. The first conversation is free and it is about your process, not our software. If automation is the wrong answer we will say so — a project that should not exist is the most expensive thing we could sell you.

AI is for companies with a budget we do not have.

That belief is the single most common reason a business does nothing, and the prices on this page are published to settle it: the smallest projects start at a hundred dollars, and the largest run past ten thousand. Tell us your budget and we will tell you what it buys — the advice is shaped around the number you actually have rather than around the largest one we could quote. If it buys nothing worth having, we will say that too, and the consultation, the design and a working demo of your own still cost nothing.

AI forgets, and it makes things up when there is too much context.

That is the exact problem TerranSoul exists to solve, and it is why the memory is data rather than weights: auditable, reversible, and answerable from retrieved passages rather than from whatever the model half-remembers. The site guide on this page is built the same way — ask it something this site does not publish and watch it decline.

It will not hold up at scale.

Scale is a shape, not a setting. The same run spreads across parallel workers, moves over interchangeable transports, and deploys inside a datacentre, on your own premises, or on a single laptop. Multi-tenant and isolation are part of the design rather than something bolted on when the second customer arrives.

It will break my production system.

Nothing we build reaches your production systems without a person releasing it. That is not a setting we switch on for nervous clients — it is the gate in the middle of the diagram, and it is the argument the company is built on. Failures take a dead-letter path where they can be inspected and replayed, rather than retried silently until something breaks.

What happens when it fails at two in the morning?

A call that fails is not lost and does not retry silently until something downstream gives way. It takes a dead-letter path where it can be inspected and replayed back through the broker — the same route it would have taken the first time. You find out what failed and re-run it, rather than discovering a week later that a day of records never arrived.

All nine, answered

Let us work out where AI actually fits.

The first conversation is free and it is about your process, not our software. If automation is the wrong answer for you, we will say so in that conversation rather than in the third invoice.

Talk to usServices and prices

Measured, not aspirational

95.0NDCG@10 · LongMemEval-S

Retrieval quality on a public 500-question long-term-memory benchmark, using hybrid keyword + dense search with reciprocal-rank fusion. Every number we publish is bound in CI to the benchmark run that produced it, so a measurement that changes makes the page fail a check instead of quietly drifting.

  • NDCG@1095.0
  • Recall@599.4%
  • Recall@20100%
  • MRR95.8
LongMemEval-S · 500 questions · hybrid FTS5 + dense + RRF

The name is the argument

Terra + Animus — a mind, given ground to stand on. The same root names our memory product, TerranSoul.

Our story

Why we exist

The real risk of AI is not that it replaces people — it is the concentration of power. Frontier models take enormous compute and capital, so a few corporations own the core technology and everyone else rents intelligence they cannot control.

Terranimus starts today to change that: own your AI — its memory and its ability to act — instead of renting it.

Two products, one platform

One remembers. One acts.

Memory

TerranSoul · the memory of AI

Long-term memory across sessions, apps, and even different models — a memory layer you own, on cloud or local AI. Solve a hard task once on a strong model; local models reuse the learned experience at almost no cost.

Execution

Tesseres · the doer

An integration and execution engine built on the Enterprise Integration Patterns — deployable from cloud to laptop, with no vendor lock-in. AI does the work; a person approves every change; every action is traceable.

What we believe

Six things we will not trade away.

  • Own it, don't rent it

    Sovereignty over your own AI — your capability, not a subscription to someone else’s.

  • Humans stay in control

    A person approves and is accountable for every AI action in the real world.

  • Memory you own

    Knowledge stays in your hands — private, portable, never locked to one provider.

  • No vendor lock-in

    Deploy anywhere — from cloud infrastructure to a single personal laptop.

  • Measured, not aspirational

    Claims are backed by public benchmarks and published research, or they are not made.

  • Value stays local

    When a business owns its AI, the value and the jobs stay in the country.

Where we are headed

Bring Sovereign AI to every individual, every business, and every nation.

Talk to us →