Session Lifecycle — Start, Name, Resume, Branch

Sessions have a lifecycle — and treating that lifecycle as a deliberate design decision produces better outcomes than treating sessions as ephemeral terminal windows. Naming, resuming, branching, and the /recap command are the mechanisms.

Quick Reference

→Start named: claude -n 'task-name' — critical for /resume navigation
→Resume: claude -c (last), claude -r (picker), claude -r 'name' (by name)
→/resume 67% faster on sessions 40MB+ since v2.1.116
→Session auto-naming (April 2026): accepting a plan in Plan Mode auto-names the session
→/recap: manually triggered summary of what was done, pending, and loaded — fires automatically after 75+ min idle
→/branch (/fork): fork the conversation — original untouched, explore on the fork
→End: auto-memory persists; conversation detail does not

Starting Sessions Right

Most developers open a terminal and type claude. This creates an unnamed session that is hard to find later. Starting with a name costs two seconds and makes every future /resume straightforward.

Session start options

Name sessions after the task or ticket

Name sessions after what they are working on: 'auth-refactor', 'JIRA-1234', 'checkout-bug', 'search-performance'. These names become meaningful when you have 20 sessions in the /resume picker and need to find the right one.

Session Auto-Naming From Plan Mode (April 2026)

When you accept a plan in Plan Mode, Claude Code automatically names the session from the plan content — using the first ~1,000 characters to generate a kebab-case name. This removes the manual naming step for task-focused sessions where you start in Plan Mode.

Resuming Sessions

Resume options

Sign in to read this article

This is a premium article. Sign in with your Google account to continue.