Our mission
Give organisations, and Britain, back ownership of their digital estate.
Agentycs exists to end an organisation's dependence on a handful of American technology companies, and on the delivery model that rents capability back to it indefinitely. What you get instead is the capability itself, running on infrastructure you hold, with privacy and long-term security as properties of the design rather than tiers of a contract.
Why we exist #
A generation of British institutions has moved the systems it depends on onto platforms owned by a handful of very large American technology companies. It was sold as modernisation, and a great deal of it was. What it also bought, quietly and almost everywhere, was dependence.
The mechanism is lock-in, and it is not subtle. Getting in is made cheap, because the purpose of the first invoice is to move the workload rather than to make money. Once your data, your identity, your tooling and your people's habits live inside a platform, the cost of leaving stops being a licence fee and becomes a re-platforming programme with a business case nobody wants to sign. That is the moment pricing power changes hands, and it does not change back.
Then there is the question of whose law applies. A platform operated by a US company is subject to US law and to the decisions of the US administration of the day, wherever the datacentre happens to stand and whatever the contract says. A government cannot fully govern a system it cannot inspect, cannot move and cannot switch off. Neither can a board.
The same shape appears on the delivery side, where consultancies are paid by the day and have no commercial reason to leave your people able to do the work without them. Capability is rented rather than transferred, the understanding walks out with the contractors, and the dependency is renewed at the next procurement. Organisations here have paid for the same capability several times over and own remarkably little of it.
Agentycs is built to invert each of those. The platform runs inside a boundary you control, your data stays in formats you can read without us, and the capability lands with your own people instead of being metered back to them. None of it depends on our continued goodwill, which is the only test of ownership that means very much.
Dependence was not a side effect of modernisation. It was the business model.
Where this goes #
A United Kingdom able to build and operate the systems that will decide the next century, no longer held back by dependence on foreign platforms or by suppliers with no reason to make themselves unnecessary.
Self-reliance here is not a nostalgic ambition, it is the precondition for building anything ambitious. A country that cannot build and operate its own advanced systems does not get to decide what those systems are for, how they behave or who they answer to. It gets to choose between other people's answers.
That means being excellent at both halves of the problem: the analytics and inference that make a capability worth having, and the engineering that makes it survive contact with a real organisation. And it means transferring the capability rather than renting it back, so that what a British institution buys once, it owns.
What we hold to #
Four commitments follow from that, and each one carries obligations we can be held to. A principle you cannot be held to is decoration.
You own it. #
Ownership is not a toggle in somebody else's console. It means the infrastructure is yours, the data is yours in a format you can read without us, the models and the weights sit on storage you control, and the running system is still yours when the commercial relationship changes. If any part of that depends on our goodwill, you do not own it.
What it commits us to
- The whole platform runs where you put it: your cloud, your datacentre, a micro server small enough to carry, or a site with no external connectivity at all.
- Model weights, datasets and credentials resolve to storage you own, so nothing the platform needs at runtime sits beyond your reach.
- Data stays in open columnar and vector formats on object storage you can point at your own bucket, readable by tools that have never heard of us.
- Tenant boundaries are derived on the server from validated identity, never inferred from a header, a hostname or a hint a client can set.
You need us for nothing. #
The real test of a sovereign system is whether it survives losing its supplier. Ours has to keep running without our network, our people and our permission, and it has to be genuinely possible to leave. We would rather be kept because we are worth keeping than because leaving was made impossible.
What it commits us to
- Remote and disconnected sites keep serving their loaded models when the core is unreachable, then reconcile when the link returns.
- The platform speaks the protocols your tools already speak: PostgreSQL wire, Arrow Flight SQL, S3-compatible storage, OpenAI-compatible inference, OIDC, SCIM, MCP and Kubernetes.
- It is built on open-source foundations and supports alternative hardware, so no single vendor sits between you and your own system.
- Your people are meant to end up able to run and extend it themselves. Capability we do not transfer is capability we are renting back to you.
Privacy and long-term security are givens. #
Neither is a tier, an add-on or a roadmap item. A system holding a country's data has to be private and defensible on the day it is installed and in ten years' time, which makes both of them design constraints rather than features. It is also why the answer, when the platform cannot resolve a rule, is no.
What it commits us to
- One authorisation decision for every privileged action, shared identically by REST, SQL, realtime and tool-protocol callers.
- Data rules travel with the data: row filters and column masks are evaluated on every read, whichever gateway asked.
- Every data access lands in a canonical, redacted audit ledger, correlated to the identity and the query behind it.
- Every build carries signed provenance and a bill of materials, and its signature is checked before anything loads.
The platform builds itself. #
Owning your own platform has always been affordable for the very largest organisations and out of reach for everyone else, because owning one meant staffing one. A software factory that writes, tests and verifies its own code changes that arithmetic, and it is the reason the other three principles are something we can offer rather than something we merely admire.
What it commits us to
- Product intent goes in as plain language; agents explore designs, plan the work in dependency order and implement it in isolated environments.
- Correctness is a property of the pipeline rather than of any one agent: validation, tests, adversarial review and live tests against actually-deployed services, every gate failing closed.
- It builds the platform it runs on, so a weakness in the factory surfaces on us long before it could surface on you.
- People decide what should exist and whether what came back is right. No approval is ever an agent's to give.
Where these commitments show up
The controls behind them, the company they came from, and the machinery that builds it.
Company
What Agentycs builds, where it operates, and what stage the company is honestly at.
Company overviewOur story
More than a dozen AI projects, one conclusion, and a deliberate break with how AI is usually delivered.
Read the storyTeam
The two founders, and how a workforce of agents writes, reviews and verifies the platform.
How the work gets doneTrust and security
How the platform enforces access, isolates tenants, verifies what it runs and stays sovereign.
See the controlsBuild on ground you own
If your organisation is deciding how much of its future to hand to somebody else's platform, we would like to talk. Bring the constraint you cannot get past — the classification boundary, the regulator, the site with no uplink, the exit clause nobody could actually execute — and we will show you how the platform handles it.