Manifesto

What we believe

Nine out of ten AI agent projects die in the pilot stage. It is not for lack of a good model. It is that almost nobody takes responsibility for the hard part: the real data, the permission, and the question that comes afterwards — why did the agent do that?

01
A demo is not a delivery
Anyone can assemble an agent that impresses in a meeting room. Few ship one that survives Monday morning. The difference is never in the model — it is in the data, in the permissions, and in who picks up the phone when something breaks.
02
Whoever does the work knows how to automate it
That person should not need a translator, nor a place in a queue. They describe in plain language what they need, and the agent exists. Making that work underneath is our problem, not theirs.
03
Automation nobody can explain is risk, not gain
Every automatic decision leaves a trail: where it came from, which version it was, who approved it. If there is no answer to “why did the agent do that?”, we do not ship it.
04
People still decide
The AI proposes, the specialist approves. Not because the machine is dumb, but because responsibility does not transfer to software.
05
Hidden cost is a bill that arrives later
Every agent shows what it consumes, by step and by area. You decide what is worth scaling before the invoice, not when it lands.
06
Nothing that is yours should be locked in
Your data, your cloud, your model. The platform runs where your policy says and swaps AI by configuration. If one day you want to leave, you leave — and you take everything with you.
07
Repetitive work does not deserve anyone's best hours
In the end, that is what this is about. Not about cutting people: about giving back the part of the work only they can do.

We are from Brasília. Our agents look after the billing of a clinic, the management data of a company, and the follow-up of malaria and tuberculosis cases in the Amazon. It is the same work: lifting the bureaucratic weight off people who have more important things to do.

Where we are