claude code (cli) 5hour limit reset via bash script or cron

claude-heartbeat-cron.md

Use a small wrapper script (see claude-heartbeat-cron.sh)

Make it executable:

chmod +x ~/.local/bin/claude-heartbeat.sh

Edit crontab:

crontab -e

Add:

0 7  * * * ~/.local/bin/claude-heartbeat-cron.sh
5 12 * * * ~/.local/bin/claude-heartbeat-cron.sh
10 17 * * * ~/.local/bin/claude-heartbeat-cron.sh

Why this is better

• No duplication risk from repeated installs
• Logging persists across reboots
• Easy to audit behavior
• Easy to remove
• Works even if your shell environment is minimal

claude-heartbeat-cron.sh

#!/usr/bin/env bash
# ~/.local/bin/claude-heartbeat-cron.sh

LOG="$HOME/.claude-heartbeat.log"

echo "[$(date '+%Y-%m-%d %H:%M:%S')] Sending Claude heartbeat..." >> "$LOG"

if claude -p "hello" >/dev/null 2>&1; then
    echo "  ✓ Claude responded" >> "$LOG"
else
    echo "  ✗ Claude failed" >> "$LOG"
fi

Claude-Heartbeat-README.md

What this dumb-ish script does

• Periodically sends a minimal request to the claude CLI

• Purpose:

  • verify connectivity
  • keep the CLI session “warm”
  • potentially advance or trigger the usage/session reset cycle if the backend treats activity as renewal

• It does not guarantee a reset of any quota or limit

• It makes no assumptions beyond “a successful request happened”

• Can be used to “pre-warm” your session: by sending small requests before heavy usage, you increase the likelihood that the first quota reset happens sooner than usual

For example:

Scenarios

You want to “pre-warm” or consume a tiny amount of quota before starting your main work.

Example:

• You start work at 09:00
• Run a small request at 06:00 via the heartbeat script to trigger the quota/timer cycle early
• After a few hours of heavy use (e.g., 11:00), the system performs its first reset, giving you a full quota refill sooner than normal

---

The cron implementation

• Performs the same heartbeat functionality but is more stable
• You create periodic pings via crontab that call and execute _claude-heartbeat-cron.sh_
• Logs each request’s success or failure to a dedicated log file
• Stateless, reliable, and minimal: each run is independent, making it easier to audit and schedule
• Ideal for automating pre-warm cycles to increase the chance of early quota resets

claude-heartbeat.sh

#!/usr/bin/env bash

# Claude CLI heartbeat
# --------------------------------------------------
# Sends a minimal request at specific times each day.
# Purpose:
#   - Confirm Claude CLI is reachable
#   - Keep the session warm
#   - Potentially refresh session/usage state
#
# IMPORTANT:
# This does NOT guarantee quota or limit resets.
# Any renewal behavior depends entirely on Claude's backend.
# --------------------------------------------------

SCHEDULED_TIMES=("07:00" "12:05" "17:10")
LOOP_INTERVAL=30   # seconds
last_run=""

echo "[claude-heartbeat] Watching times: ${SCHEDULED_TIMES[*]}"

while true; do
    now_time="$(date '+%H:%M')"
    today="$(date '+%Y-%m-%d')"

    for t in "${SCHEDULED_TIMES[@]}"; do
        key="${today} ${t}"

        if [[ "$now_time" == "$t" && "$last_run" != "$key" ]]; then
            last_run="$key"
            echo "[$(date '+%Y-%m-%d %H:%M:%S')] Sending Claude heartbeat..."

            if claude -p "hello" >/dev/null 2>&1; then
                echo "  ✓ Claude responded successfully"
            else
                echo "  ✗ Claude CLI failed"
            fi
        fi
    done

    sleep "$LOOP_INTERVAL"
done


### Notes

# Time matching is **minute-accurate**, not second-accurate
# Loop interval can drift slightly, but the `last_run` guard prevents double-fires
# If the machine sleeps at the exact trigger time, the event is missed