Run agents
One command drives every coding CLI you already pay for — the same flags everywhere.
Run
The full flag reference for driving any CLI headless or interactive.
Pin versions
Lock each CLI to a version per project with agents.yaml.
Rotate accounts
--rotate spreads runs across every login so no one account rate-limits.
Schedule
Wrap any run in a cron entry for nightly audits and recurring triage.
Monitors
Watch a source, detect a change, fire an agent — event-triggered instead of clock-triggered.
PTY
Drive interactive terminals, REPLs, and TUIs from an agent.
Computer
Let an agent screenshot, click, and type in native desktop apps.
Next
Running one agent is the on-ramp, not the destination. Once a single run works, the same harness scales two ways: a team of agents splits one task across several isolated workers with boundary contracts, and Schedule / Monitors keep an agent running unattended — on a cron, or reacting to an event — with the same Keychain-backed secrets scoping it would use interactively.