Activity

1. opencode-agent commented on Oct 21, 2026

   

   opencode-agentBot

   on Oct 21, 2026 – with OpenCode Agent

   Contributor

   More actions

   Thanks for the request! If I understand it right, you'd like the agent in v2 to keep an eye on a process it started: notice when it fails, fix the problem and rerun it, as it could in v1 by sleeping and checking the output.

   Some context on how v2 works now: sleep isn't blocked. When a shell command runs in the background, the agent is told not to poll it with sleep/tail. It gets a notification with the command's output automatically when the command finishes, and can then fix and rerun. That covers commands that finish. It doesn't cover watching a process while it is still running (dev servers, watchers, long builds), which I think is the gap you're describing. The polling rule applies "unless the user explicitly asks otherwise", so for now you can tell the agent directly to check the process's output periodically.

   Related issues and PRs:

   • feat: add monitor events #52569: open PR adding a monitor tool that streams a running process's output into the session and wakes the agent
   • Proposal: background subagents, monitors, cron, worktree isolation #49842: proposal including monitors for background processes
   • [FEATURE]: Let agents wait and automatically continue a session #53960: request to let agents wait and continue a session automatically
   • fix(core): tighten background shell polling guidance #51436: PR about the background polling guidance

   Maintainers will take a look and consider it.
2. borjaBruker commented on Oct 21, 2026

   

   borjaBruker

   on Oct 21, 2026

   Author

   More actions

   I have managed to make it work, after using a smarter model, and insisting, and pasting this response.
   Still feels like a regression, that it rejects to poll a process on first attempt.