The first thing I tried when testing Antigravity (coming from Claude code agent teams) was /teamwork-preview. It seems neat, but it spawns a project orchestrator subagent, and then that subagent spawn grand-children agents to do the actual work.
It seems crazy to not be able to see or browse into these subagents. The /agents command only shows me my immediate child, and I cannot recursively browse the tree. Likewise the IDE seems to only show my direct ancestors in the "Subagents" view. Of course you can ASK the agent to try to summarize what they're doing, but that does not feel sufficient. I'm not even sure they're correctly created.
This seems like a major gap. In Claude code I can just press the down arrow and switch my view to any of the running agents. Likewise, other agent orchestrators I've used (that sit above the claude/codex/opencode level) all naturally provide a way to drill down and see what individual agents are doing.
The first thing I tried when testing Antigravity (coming from Claude code agent teams) was
/teamwork-preview. It seems neat, but it spawns a project orchestrator subagent, and then that subagent spawn grand-children agents to do the actual work.It seems crazy to not be able to see or browse into these subagents. The
/agentscommand only shows me my immediate child, and I cannot recursively browse the tree. Likewise the IDE seems to only show my direct ancestors in the "Subagents" view. Of course you can ASK the agent to try to summarize what they're doing, but that does not feel sufficient. I'm not even sure they're correctly created.This seems like a major gap. In Claude code I can just press the down arrow and switch my view to any of the running agents. Likewise, other agent orchestrators I've used (that sit above the claude/codex/opencode level) all naturally provide a way to drill down and see what individual agents are doing.