Continue vs Resume: When Each Matters

+15 Mana ✨
CLI Reference: Resume and Continue(DOCUMENTATION)

Introduction

Hermes has two ways to pick up an old session: --continue (the most recent one) and --resume <id> (a specific one by ID). They look interchangeable in a one-session-a-day workflow but become very different at scale. Knowing which to reach for is the difference between "always working in yesterday's session" and "always working in the right session."

Key Concepts

  • --continue (or -c): Picks up the most recent session, whichever you touched last.
  • --resume <id>: Picks up a specific session by its ID.
  • --resume "<title>": Picks up a specific session by its title.
  • hermes sessions list: Prints your past sessions to the terminal, IDs, titles, and recency. The starting point for resuming an older session whose ID you do not remember.

Real World Context

A developer running one Hermes session at a time always uses --continue, there is only ever one "most recent." A developer juggling four parallel sessions (a refactor, a bug, an investigation, a doc draft) uses --resume <id> for three of them and --continue for the one they were last in. Same product, two workflows.

Deep Dive

Side-by-side

bash
hermes --continue                # last touched session, no thinking required
hermes -c                        # short form
hermes -c "my refactor"          # named session, latest in its lineage

hermes --resume abc123def        # specific session, by ID
hermes --resume "my refactor"    # specific session, by title
hermes sessions list             # browse all past sessions from the CLI

When each wins

text
Situation                                Reach for
──────────────────────────────────────   ─────────────────
One active task                          --continue
Many parallel sessions                   --resume <id>
Resuming yesterday's task                --continue if it was last
Resuming a week-old task                 --resume <id>

How to find session IDs

Run hermes sessions list to print your sessions to the terminal, sorted by recency, with IDs and titles. Once you spot the session you want, resume it with hermes --resume <id> or, more conveniently, hermes --resume "<title>".

Common Pitfalls

  1. Always typing --continue, Convenient until you accidentally branch the wrong session. If you have multiple active sessions, an explicit --resume is safer.
  2. Forgetting that titles work too, --resume "<title>" is often easier than tracking down a session ID. If the title is unique, resume by title.

Best Practices

  1. Title sessions immediately, A meaningful title is what makes --resume "<title>" and hermes sessions list useful later.
  2. Use -c for the dominant session, the list for the rest, Keep one "hot" session that --continue always lands on; use hermes sessions list + --resume "<title>" for everything else.

Summary

  • --continue resumes the most recent session; --resume <id> or --resume "<title>" resumes a specific one.
  • hermes sessions list is the CLI way to browse past sessions and find the ID or title to resume.
  • One-session workflows live on --continue; many-session workflows live on sessions list + --resume.
  • Title your sessions or you will lose them in the haystack.

Code Examples

bash
# The dominant session, the one you keep going back to today
hermes -c

# Browse past sessions on the CLI to find the one you want
hermes sessions list

# Pinpoint resume, by ID or by title
hermes --resume 7f3a9b2c1e
hermes --resume "refactor-auth-middleware"
✓ Completed