Shortlists with proof, then outreach, in one click.

How it works

Evidence-backed matching, then outreach

The part that usually eats an afternoon. One click reads the JD, scores your pool against it with quoted proof, and drafts the messages for the people who actually fit.

1. 1

   The role becomes atomic requirements

   Must-haves, nice-to-haves and disqualifiers are split one-per-item with retrieval keywords attached. Must-have means required to hire; disqualifier means deal-breaker.

2. 2

   The pool narrows before any scoring

   Up to 200 candidates are considered and the obvious misses are dropped cheaply, so model time is spent on people with a real chance.

3. 3

   Each candidate is read, one at a time

   Up to 25 candidates get a full pass. Every must-have gets met, partial or not_found — and every met claim has to carry a near-exact quote lifted from the resume.

4. 4

   Quotes are checked against the resume

   Evidence verification re-checks the model's quotes against the actual resume text. A claim with no support in the document does not survive into the tier.

5. 5

   Perfect, Partial, Rejected

   Tiers come from verified evidence, not a vibe score — so you can open any candidate and see exactly which requirement the decision rested on.

6. 6

   Perfect fits get drafted messages

   Email and LinkedIn drafts are written for the top fits, aimed at the goal you picked — screen, submit, or a rate conversation — in your recruiter voice.

7. 7

   The playbook learns your desk

   Daily notes distil into a playbook that is injected back into JD parsing and scoring, so the agent gradually matches how your team actually screens.