Run engagement
Parallel agents · provider failover · N-model validator voting.
Targets (one URL per line)
Model panel (1st = primary · others fail over & vote)
Validator votes (N)
Max agents (0 = all)
Offline (pipeline self-test)
Use subscription (Claude/Codex login)
▶ Run harness
Agent library
…
Models
OpenAI-compatible providers — CLI & API.
Report
Last engagement.
Run an engagement to generate a report.