Repository navigation
Can't chat with the subagent session anymore v1.1.1 #6907
Description
Question
I heavily use the multi agents capabilities. During the time of execution. I tend to have conversation with the subagents session instead of the primary agent session. In order to save the session from being bloated, and sometimes queue the messages in the subagent session when the agents go sideways. It was very handy.
Is it a bug that we don't see the chat in subagents?
Metadata
Metadata
Assignees
Labels
No labels
Activity
rekram1-node commentedon Jan 5, 2026
This was intentional, tho note that even before it got removed, you werent actually talking to subagent it was still primary agent just w/ the context of subagent
anjanpoonacha commentedon Jan 5, 2026
oh oh!!
I slightly altered the question to reflect what I really meant.
Sad to see the powerful feature sunset!!
bobab12 commentedon Jan 5, 2026
This was a critical part of my workflow as well. Being able to send messages to subagent sessions allowed me to course-correct when needed and significantly improved subagent performance. Would love to see this restored.
Re: routing to the primary agent - that doesn't quite match my experience. When I sent messages while viewing a subagent (e.g., telling it to use
ghCLI instead of webfetch for investigating source code), the subagent changed its behavior directly. Could be wrong, but it felt like direct communication. Either way, the functionality was valuable for steering subagents.Edit: did some digging and you're technically correct that the TUI agent config was used, not the subagent's. But messages did go to the child session and became part of its context, which influenced subsequent behavior. That's what made it useful. Hope it can come back in some form.
OmarGoubail commentedon Jan 5, 2026
I used this constantly too, It was very useful.
vladislavlozhkin commentedon Jan 5, 2026
This feature was exactly the reason I switched from Claude Code to OpenCode. The ability to interact with subagent sessions and course-correct them mid-execution was a game-changer for my workflow. Really disappointing to see it removed — might as well switch back to Claude Code now.
anjanpoonacha commentedon Jan 6, 2026
In fact, I stopped using the Opencode Desktop because this specific feature was removed.
I had created a question for that too! #6191
gund commentedon Mar 5, 2026
Subagent steering is very important as if they start to deviate you currently have no other option but to terminate the entire MAIN agent session, even when it may run multiple subagents in parallel, and because of one subagent session you would waste the useful job done by other subagents. Its simply very inefficient.
tbtmuse commentedon Mar 5, 2026
Opensource is wonderful, I forked it, and put the feature back, you can do the same, happy coding 😎
anjanpoonacha commentedon Mar 5, 2026
💯
I did one for myself
danrayson commentedon Mar 6, 2026
Part of the workflow I've been making required subagents ask the user clarification while they were running. It would help massively on course correction and intervention. Either the subagent asking questions when it needs to or being able to hulksmash escape and interject would, I would think, be a critical feature.
I was about to suggest it as a feature, but saw this thread.
EDIT: I just found out that the subagent will kick you out of the subagent flow back to the primary agent and ask its questions there instead, once you've answered it goes back into the previous task it was running. Perhaps I was expecting the worst, but I got the best :)
github-actions commentedon May 6, 2026
To stay organized issues are automatically closed after 60 days of no activity. If the issue is still relevant please open a new one.