Open-source coding agent CLI. Switch models in one session and finish explicit goals against checks you approve.
Terminal score: 0–1000 raw, weighted across 4 dimensions. Public score: 0–10 normalized (shown in the 30-day stars chart above).
