pinecall kill

pinecall kill [agent]                    … and with --prod, in production

An admin's verb: the agent let go by every app that holds it in the environment, whoever's process it is — a colleague's pinecall start, a deploy, the org's own server. Each app is sent an error coded killed on that agent (POST /v1/agents/{slug}/kill, a key that opens both team and app): a pinecall start holding it prints <agent> ✗ <agent> was killed by <name>, and every app keeps answering for its other agents and does not declare this one again when its socket comes back. A pinecall start is two apps — the agent's process and the CLI's own socket beside it, which answers the console — so one kill lets go of both, and the process is the one that prints. The agent's live calls move as on a drain, to another holder or parked. From that moment it is in no agent list, no console and no playground.

Nothing is erased: its settings and their versions, its calls and their logs stay, and an app that declares it again — a pinecall start restarted, a save that reloads the class — holds it again. Without agent, the agent beside the class; from any folder, the slug named. A key that is not an admin's is refused with exit 2; an agent no app holds in the env is the gateway's 404, exit 2.

$ pinecall kill front-desk
killed front-desk in sandbox: 2 apps let go of it (app_7, app_8) and kept their other agents — its settings and calls stay, and an app that declares it again holds it again