## Optimize your setup

One-click toggles to tune Power Claude for how you work.

Optional, one-click optimizations. Each button flips a real setting — turn on what fits how you work; everything is reversible from the dashboard or VS Code Settings. (In the terminal, run the matching command from the dashboard or settings instead.)

- Token Saver — trim low-signal noise (logs, command output, diffs, JSON) out of what Claude reads before it's forwarded, so more of the capacity you already pay for goes to real work. On by default; set how aggressive from the Token Saver panel.
- Warm Compacting — skip Claude Code's long compaction pause: Power Claude pre-summarizes the earlier conversation in the background (sidecar-only, never in your live context) and swaps it in near the wall. On by default (free, no license required); resilient fallback, so it silently falls back to normal compaction if ever unsafe.
- Auto-resume sessions — let Claude Code continue through stalls and rate-limit waits instead of waiting on you.
- Detached Sessions (experimental · Pro) — keep a turn running when your browser tab sleeps, refreshes, or you close the window; the dashboard reattaches and replays what you missed.

> 💡 **Tip:** Detached Sessions needs Linux with a systemd --user manager; on other hosts the toggle stays disabled. It's experimental — leave it off until you've tried it on a session you can watch.

[Open Token Saver](command:powerClaude.tokenSaver.open) · [Toggle auto-resume](command:powerClaude.toggleAutoResumeSessions) · [Toggle Detached Sessions (Pro)](command:powerClaude.toggleDetachedSessions) · [Open the dashboard](command:powerClaude.dashboardV2.focus)

[Tuning & optimization →](https://neural-llm.com/docs/power-claude#optimize)