Skip to main content
Every door runs the same loop — seed a digital twin, run your agent, score the tape — and lands on the same first scored report. Pick the one that matches how you work.

Claude Code

Paste one prompt; your Claude Code session drives the whole run.

I already have an agent

Point Pome at your own repo and register it for scored runs.
More doors (Cursor, CLI) are on the way — each reaches this same report.