We built the tool we needed.
AgentR started as a complaint, not a company. It is still mostly a complaint, just one with an evaluation pipeline attached.
We once spent three weekends reading two hundred resumes for one backend role. Half of them claimed to have led the migration. Two of them actually had. By Sunday afternoon of the third weekend nobody was reading carefully any more, and we all knew it.
The tools we tried made it worse. They ranked people by keyword bingo, which meant the candidates who had read a blog post about resume optimisation beat the candidates who had done the work. That is not a filter. That is a lottery with extra steps.
So we built the thing we wished existed. It reads every application properly, against criteria you write down. It checks what people claim against the public record. It shows its work, every time, with the line of the resume that earned the score attached.
You still make every decision. You just stop doing the paperwork.
Three things we refuse to move on.
- 01Evidence beats polish
A language model can make any career sound senior in nine seconds. It cannot invent an outcome that never happened. So a bare skills list earns nothing, and a result with a number still attached earns a lot.
- 02The human decides, always
Nothing is auto-rejected. The whole pile reaches you, ordered, weak end included. Where to stop reading is a judgement call, and judgement calls belong to people.
- 03Show the working
Every score carries the line that earned it. Every flagged claim carries its sources. If we cannot show you why, we do not get to assert it.
Where we are right now.
Private beta
We onboard teams in the order people signed up, slowly enough to sit with each one properly.
Small on purpose
Small enough that the person who answers your email is the person who can change the product.
Talking to recruiters
If you hire and you have opinions about screening, we would genuinely like to hear them. Write to us.