The control layer for AI
Gluon prevents wrong answers and wrong actions, so every organisation can explain every decision its AI takes.

Every organisation runs on decisions it already took: which version is in force, who may see what, who signs above which amount. AI knows none of them. It finds what looks right and answers with confidence. That gap is why Gluon exists.
An ordinary day
Principles
They decide what we build, and what we refuse to build.
- 01Your data is yours. Your models are replaceable. So are we.We build so you can leave. That is how you know you stay for the right reason.
- 02We’d rather say “we don’t know” than invent an answer.A gap that is shown is safer than a guess that sounds right.
- 03Relevant is not the same as applicable.Similar text is not a rule. Status, version and scope decide.
- 04AI should know when to stop.Some actions need a person. Your rules decide which, not the model.
- 05You stay because it works. Never because you can’t leave.Open code, your own environment and no separate edition.
- 06Whatever decides what AI may do should be open to read.The rules that govern AI belong in public code, not in a black box.
From answers you can trust to AI you can hold to account.
Every step is built as open source.
AnswersNo wrong answers.AI gets only what applies, within the permissions of the person asking, with a decision trail for every answer.
ActionsNo wrong actions.Before an agent sends, pays or changes anything, Gluon decides: allowed, approval required or blocked.
AccountabilityAI you can hold to account.Every organisation can explain every decision its AI takes, to its board, its regulator and itself. And can still switch any of it off.
In October 2026 Gluon Desktop, Drift, Gluon Connect’s new sources and the document inspector landed.In pilot at YallaSwap, Lalaaji, Tuli, inbox2 and Servaid; their results are theirs to tell.

Built in the Netherlands
We build with a small number of design partners. How they work sets our priorities. Shaped by the people who use it.
The NetherlandsWhere we build.
EuropeWhere Gluon runs.
- Your environmentWhere your data stays.
- Open sourceWhat decides what AI may do is open to read.
Design partners
Design partners work directly with the team that builds Gluon: on their own sources, with their own rules, around one concrete use case.
| You bring | You get | |
|---|---|---|
| /01 | One use case | A working use case |
| /02 | Your sources | A direct line to the builders |
| /03 | Your rules | A say in the roadmap |
| /04 | Your users | The code, without lock-in |
Start with
one use case.
- Choose one question and its outcome.With the people who build Gluon, you pick one question that must be right, with an owner and a clear outcome.
- Set the boundary and connect the sources.Name the sources, their owners and what must never be returned. Gluon connects them in your own environment, with the permissions they already have.
- Test it with your people.Your people ask real questions, including conflicting versions and requests outside their scope. Their questions and corrections show what to refine.
- Decide and go live.Business, security and operations judge the result against the outcome you set. Then the use case runs where your people work.
Contact
- [email protected]
- Company
- Sadiq Beheer B.V.
- Address
- Coolhaven 238a, Rotterdam


