Evidence over claims
If a statement about a model cannot be re-run from a bundle, we do not make it. That applies to our marketing as much as to our packs.
Company
Built for the teams accountable for private AI.
Coyos builds Etalon for organisations that run language models inside their own networks and have to answer for what those models are allowed to do. Risk, security and platform teams, usually where “we tried it and it seemed fine” is not an acceptable answer.
About Coyos
Small, engineering-led, and specific.
Coyos d.o.o. is the company behind Etalon. We started with the narrowest useful problem we could find: a private classifier in front of a support-desk router, and the evidence a team needs before letting it route real contacts. The runner is open source. The packs are how we make a living. Both are described plainly on this site.
An étalon is a measurement standard: the reference you check other instruments against. That is the layer we build for private AI deployments. Not the interface, not the model, but the evidence a reviewer can check later.
We are based in Europe and work with customers whose deployments are on-prem, in a VPC, or air-gapped.
Principles
Four rules we apply to ourselves.
Nothing phones home
The runner has no telemetry, no licence check and no upload. This site has no third-party scripts. Your data is never ours to see.
The runner is open
The code that touches your endpoint and your data is public and Apache-2.0. If this site and the repository ever disagree, the repository wins.
You make the call
Etalon produces a scoped decision with reasons. What it means for your deployment, your contracts and your obligations is your judgement, not ours.
Contact
Talk to us.
Tell us what your model does in production and where it runs. You will get a plain answer about which packs your reviewers will ask for, or whether Etalon is the wrong tool for it.
- hello@coyos.ai
- Runner
- github.com/coyos-ai/etalon
- Company
- Coyos d.o.o.
Request a demo or pack accessOpens your mail client