Do you need our model or code?

No. We run the agent as you ship it.

Will you publish our results?

Only anonymised aggregates, and only with your approval. NDAs are available.

What happens to our data?

It’s used only for your evaluation and deleted when the programme ends, unless you ask us to keep it. Our privacy policy covers what this website stores; the programme agreement covers the evaluation data.

What if our agent does badly?

That’s useful to know, and it stays between us. Nothing about an individual partner is published without approval, and the programme exists to find these gaps before your customers do.

Teams that run agents on real tasks and want an independent measure of how often “done” is true. If you’re not sure the programme fits, contact us with a short description of what your agent does.

Design partners

Become an Inoku design partner

Apply to the design-partner programme: early evaluations of your agent's completion claims, free during the programme, for teams running real agents.

We’re working with a small group of teams that build or deploy AI agents on real systems, such as coding, operations and data agents. Together we’ll measure how often your agent’s reports match what actually happened.

What you get

Early evaluations of your agent’s completion claims, free during the programme.
A private report, broken down by task type.
A say in what the benchmark measures and how results are reported.
A first look at our open-source tools.

What we ask

A set of real or realistic tasks your agent handles.
The ability to run your agent as it is: no model weights, no code changes.
A 30-minute call every two weeks.
Permission to publish anonymised, aggregate findings. You can opt out.

Timeline

Applications are open now. The programme runs from Q1 to Q2 2027, with the first evaluations starting in Q1.

We review applications as they arrive and reply to every one. If the programme is full, we’ll say so and keep your application for the next round if you agree.

Design-partner application

Apply

About five minutes. Five places; we reply to every application and say which round you are in.

What happens next

We read your application and reply by email, usually within a few working days.
A 30-minute call to agree the task families and how results are reported.
The first evaluation of your agent, as you ship it, at no cost during the programme.

Questions

Frequently asked questions

Is it free?: Yes, during the programme.
Do you need our model or code?: No. We run the agent as you ship it.
Will you publish our results?: Only anonymised aggregates, and only with your approval. NDAs are available.
What happens to our data?: It’s used only for your evaluation and deleted when the programme ends, unless you ask us to keep it. Our privacy policy covers what this website stores; the programme agreement covers the evaluation data.
What if our agent does badly?: That’s useful to know, and it stays between us. Nothing about an individual partner is published without approval, and the programme exists to find these gaps before your customers do.
Who should apply?: Teams that run agents on real tasks and want an independent measure of how often “done” is true. If you’re not sure the programme fits, contact us with a short description of what your agent does.