Early access is openRequest access
Mowa

MOWA FOR DOMAIN EXPERTS

Your judgment, made runnable.

You know what a good answer looks like. Mowa turns that knowledge into scorers the whole team can run on every change, so quality stops depending on you being in the room.

RECOMMENDED

Scorers suggested from the prompt itself.

Mowa reads a prompt and recommends the scorers it deserves. You start from a drafted rubric that already understands what the prompt is trying to do, instead of a blank page.

  • AI-recommended scorers drafted from the prompt
  • Edit the fields and thresholds to match your standard
  • Adopt, adjust, or discard each suggestion
scorer/policy-tonerecommended

source: prompt support-triage v13

fields: tone 1-5 · policy_pass bool

status: draft, awaiting your edit

DRY RUN

Test a scorer on real rows first.

Before a scorer judges anything that matters, run it against dataset rows and read how it grades cases where you already know the right answer. Calibrate it until it agrees with you.

  • Run any scorer against individual dataset rows
  • Compare its verdicts with your own
  • Adjust the score interface until it matches your judgment
scorer/policy-tone · dry runrow 17

your call: tone 4 · pass

scorer call: tone 4 · pass

agreement holding across 12 rows

SCORERS

Turn what you know into a scorer.

Build scorers in free categories with custom score interfaces: the fields, scales, and labels that match how you actually judge output. Once defined, your judgment runs on every experiment, not just when you have time to review.

  • Free categories, no fixed rubric
  • Custom score interfaces with your own fields
  • Runs on every experiment automatically
  • Scorers and experiments are included in Pro

scorer defined once applied to every run since

EDGE CASES

Datasets that cover what you worry about.

Import real cases from CSV or generate synthetic rows with guidance, so the edge cases living in your head become rows in a dataset that every future change gets tested against.

  • Dataset shapes define what a case looks like
  • CSV import for the cases you already have
  • Guided synthetic generation for the long tail
dataset/kyc-edge-cases148 rows

96 imported from csv

52 synthetic, guidance: refund + hold

shape: ticket, locale, expected

How it runs

one loop, every record kept

01 prompts

versioned, reviewed

02 datasets

rows + synthetic

03 runs

versions × models

04 scorers

your judgment

05 experiments

pass rates

MOWA FOR DOMAIN EXPERTS

Put your expertise on every run.

Start free and build your first dataset from the cases you know best. Pro unlocks scorers, experiments, and the playground. Scale when you are ready.