Small, independent, and not pretending otherwise.
Marbl3 is an IT consultancy. We work on one thing: the systems behind agentic AI and the platform engineering underneath them. We are early enough that our own platform is the best evidence we have.
An honest description.
We are a handful of people, not a firm with a bench. There is no industry practice, no managed-service arm and no vendor partner tier, which means there is nothing pushing us to widen a scope that does not need widening.
We have one client: UON. That is the whole list. We would rather write that than fill a page with logos we have no right to use, and it is worth knowing before you hire us, as is the fact that we will tell you when a problem needs a bigger firm.
What we do have is a platform we built and run. an agent platform written in Rust: a domain core that depends only on traits, capability-scoped WASM plugins, a workflow engine as the single execution path, and a LAN job mesh that spreads work across whatever machines are on the network. It is how we deliver, and reading it will tell you more about how we think than any capability deck.
Every engagement leaves architecture decision records, runbooks and a test pipeline in your repository, under your licence. An engagement that leaves you dependent on us has failed, however well it went.
Four positions we will argue for.
Not values on a wall. These are the arguments we have in engagements, and how we resolve them.
Production or it did not happen
A pilot that never took real traffic taught you nothing a whiteboard could not have.
Small, and staying small
We are deliberately a handful of people. There is no bench to keep busy, so there is no incentive to widen a scope that does not need widening.
We write down the trade-off
Every architectural decision ships with its alternatives, its costs, and the conditions under which we would reverse it.
Hand-over is the deliverable
An engagement that leaves you dependent on us has failed, however well it went.
The same four phases, every time.
United States and Europe · working across Pacific to Central European hours
Read the terrain
A short, fixed-price read: the code, the constraints and the people who carry the pager. We write down what we find, plainly, including the parts you will not enjoy.
Prove the thin slice
One real path, end to end, in production, instrumented. A thin slice settles arguments that a business case cannot.
Industrialise
Harden the slice: contracts, evaluation gates, observability, runbooks. This is where most programmes stop and where the value actually accrues.
Hand over the keys
Your engineers have been in the rotation from the start. We leave with the documentation written, the tests green and the deployment path transferred.
Tell us what is stuck.
Send a short note about the system, the constraint and the deadline. A partner replies within two working days, and if we are not the right firm for it we will say so and point you somewhere better.