Subagent History
Read what the children Claude Code and Codex spawned actually did — their conversation, tool calls and results — after the delegation has finished.
Delegation used to be a spinner and a summary. A child agent would spin up, do real work, then disappear — leaving you with whatever the parent chose to tell you about it. When the summary was thin or wrong, there was nothing to check it against.
Subagent history keeps those children readable.
Find the count
When Claude Code or Codex spawns children, a count appears beside the delegation badge on that terminal. It opens a searchable list of every subagent from that run.
Finished children stay in the list. A delegation is still readable after its spinner has gone.
Read one
Pick a subagent and you get:
- its conversation
- its tool calls
- its results
What it actually did — not the parent's summary of it.
Follow a running child
| Control | What it does |
|---|---|
| Follow log | Streams a child that is still working |
| Earlier logs | Pages back through a long run |
| Latest | Jumps to the end |
Carry a finding back
- Copy ID — the subagent's identifier, for reference.
- Copy follow-up for parent — a ready-made prompt to paste into the parent conversation, so acting on something you spotted in a child does not mean retyping it.
Read-only by design
1DevTool never starts, writes to, or steers a native child. The panel only reads.
That is deliberate: a subagent that two things are driving is a subagent whose transcript no longer explains its own behaviour. Reading is safe; steering would make the record you came here for untrustworthy.