Thread view
The thread view splits the terminal into three panes:
- Stream (top):
axel thread stream. Messages render Markdown; tool rows (▸) open the call input and output from the Claude Code transcript. - Prompt (bottom left): over
axel thread compose.enteradds a line,ctrl+entersends.↑on the first row walks this thread’s history. - Cost (bottom right): tokens, cache hits and cost of the thread, per agent.
Header
Section titled “Header”The top row shows a card per open thread: title, a wave while agents work, ● for unread.
tab opens the header, ← → picks a thread, enter switches.
Agents in a thread
Section titled “Agents in a thread”Type @name in the prompt, then ← → on the name:
- + add: axel summons the agent into the thread.
- ⇄ replace: when the current agent’s turn ends, codehead stops its session and starts the new agent with the prompt and a handoff file (task, recent conversation, changed files).
Branching a thread
Section titled “Branching a thread”The > command branch copies the thread into a new thread in the same directory, with a
fork of each agent’s Claude Code conversation. The old thread stays.
When the view cannot open, codehead falls back to axel thread open. Errors go to
~/.axel/logs/codehead-view.log.