diff --git a/watchdog/main_pi/watchdog.py b/watchdog/main_pi/watchdog.py
new file mode 100644
index 00000000..34b96b39
--- /dev/null
+++ b/watchdog/main_pi/watchdog.py
@@ -0,0 +1,199 @@
+#!/usr/bin/env python3
+"""
+Watchdog for the main Pi that pings the Pi Zero and power-cycles its relay
+after repeated failures, with cooldown and daily limits.
+
+Cron setup (every 10 minutes at :05):
+  1) Edit crontab:  crontab -e
+  2) Add the line:
+     5,15,25,35,45,55 * * * * /usr/bin/python3 /home/pi/pyro-engine/watchdog/main_pi/watchdog.py >> /home/pi/watchdog_main.log 2>&1
+
+Adjust the paths to match where this repo lives on the Pi.
+"""
+
+import datetime as dt
+import logging
+import subprocess
+import time
+from dataclasses import dataclass
+from pathlib import Path
+
+import RPi.GPIO as GPIO
+
+# ================= CONFIG =================
+
+RELAY_PIZERO = 16
+PIZERO_IP = "192.168.1.98"
+
+PING_COUNT = 2
+TIMEOUT = 2  # seconds, used for ping timeout
+
+MAX_FAILS = 3
+POWER_OFF_TIME = 15
+
+COOLDOWN_SECONDS = 30 * 60
+MAX_REBOOTS_PER_DAY = 3
+
+STATE_DIR = Path("/tmp")
+LOG_FILE = Path("/home/pi/watchdog_main.log")
+
+FAIL_PIZERO_FILE = STATE_DIR / "fail_pizero"
+LAST_REBOOT_FILE = STATE_DIR / "last_reboot_pizero"
+DAILY_REBOOT_FILE = STATE_DIR / "daily_reboots_pizero"
+
+# ================ LOGGING =================
+
+logging.basicConfig(
+    filename=str(LOG_FILE),
+    level=logging.INFO,
+    format="%(asctime)s | %(levelname)s | %(message)s",
+    datefmt="%Y-%m-%d %H:%M:%S",
+)
+
+# ================ GPIO ====================
+
+GPIO.setmode(GPIO.BCM)
+GPIO.setwarnings(False)
+
+GPIO.setup(RELAY_PIZERO, GPIO.OUT, initial=GPIO.HIGH)
+
+# ================ IO HELPERS ==============
+
+def read_int(path: Path, default: int = 0) -> int:
+    try:
+        return int(path.read_text().strip())
+    except Exception:
+        return default
+
+
+def write_int(path: Path, value: int) -> None:
+    path.write_text(str(value))
+
+
+def read_text(path: Path, default: str = "") -> str:
+    try:
+        return path.read_text().strip()
+    except Exception:
+        return default
+
+
+def write_text(path: Path, value: str) -> None:
+    path.write_text(value)
+
+# ================ CHECKS ==================
+
+def ping_host(ip: str) -> bool:
+    try:
+        subprocess.check_output(
+            ["ping", "-c", str(PING_COUNT), "-W", str(TIMEOUT), ip],
+            stderr=subprocess.DEVNULL,
+        )
+        return True
+    except subprocess.CalledProcessError:
+        return False
+
+# ================ FAIL COUNTERS ===========
+
+def update_fail_counter(ok: bool, fail_file: Path, label: str) -> int:
+    if ok:
+        logging.info("%s check OK", label)
+        write_int(fail_file, 0)
+        return 0
+
+    fails = read_int(fail_file, 0) + 1
+    logging.warning("%s check FAILED (%s/%s)", label, fails, MAX_FAILS)
+    write_int(fail_file, fails)
+    return fails
+
+# ================ REBOOT GUARD ============
+
+@dataclass(frozen=True)
+class RebootGuard:
+    cooldown_seconds: int
+    max_reboots_per_day: int
+
+    def _read_daily(self, daily_file: Path) -> tuple[str, int]:
+        today = dt.date.today().isoformat()
+        raw = read_text(daily_file, "")
+
+        if not raw:
+            return today, 0
+
+        parts = raw.split()
+        if len(parts) != 2:
+            return today, 0
+
+        day, count_s = parts[0], parts[1]
+        try:
+            count = int(count_s)
+        except Exception:
+            count = 0
+
+        if day != today:
+            return today, 0
+
+        return day, count
+
+    def can_reboot(self, now_ts: int, last_reboot_file: Path, daily_file: Path, label: str) -> bool:
+        last_ts = read_int(last_reboot_file, 0)
+        if last_ts and (now_ts - last_ts) < self.cooldown_seconds:
+            remaining = self.cooldown_seconds - (now_ts - last_ts)
+            logging.warning("%s: reboot blocked by cooldown, remaining_seconds=%s", label, remaining)
+            return False
+
+        day, count = self._read_daily(daily_file)
+        if count >= self.max_reboots_per_day:
+            logging.warning(
+                "%s: reboot blocked by daily limit, count=%s, limit=%s",
+                label,
+                count,
+                self.max_reboots_per_day,
+            )
+            write_text(daily_file, f"{day} {count}")
+            return False
+
+        return True
+
+    def record_reboot(self, now_ts: int, last_reboot_file: Path, daily_file: Path) -> None:
+        write_int(last_reboot_file, now_ts)
+
+        today = dt.date.today().isoformat()
+        day, count = self._read_daily(daily_file)
+        if day != today:
+            day, count = today, 0
+
+        count += 1
+        write_text(daily_file, f"{day} {count}")
+
+def power_cycle(relay_gpio: int, label: str, last_file: Path, daily_file: Path, guard: RebootGuard) -> None:
+    now_ts = int(time.time())
+
+    if not guard.can_reboot(now_ts, last_file, daily_file, label):
+        return
+
+    logging.warning("%s: power cycle triggered", label)
+    GPIO.output(relay_gpio, GPIO.LOW)
+    time.sleep(POWER_OFF_TIME)
+    GPIO.output(relay_gpio, GPIO.HIGH)
+    logging.info("%s: power restored", label)
+
+    guard.record_reboot(now_ts, last_file, daily_file)
+
+# ================= MAIN ===================
+
+def main() -> None:
+    guard = RebootGuard(
+        cooldown_seconds=COOLDOWN_SECONDS,
+        max_reboots_per_day=MAX_REBOOTS_PER_DAY,
+    )
+
+    fails = update_fail_counter(ping_host(PIZERO_IP), FAIL_PIZERO_FILE, "Pi Zero")
+    if fails >= MAX_FAILS:
+        power_cycle(RELAY_PIZERO, "Pi Zero", LAST_REBOOT_FILE, DAILY_REBOOT_FILE, guard)
+        write_int(FAIL_PIZERO_FILE, 0)
+
+if __name__ == "__main__":
+    try:
+        main()
+    finally:
+        GPIO.cleanup()
diff --git a/watchdog/pi_zero/watchdog.py b/watchdog/pi_zero/watchdog.py
new file mode 100644
index 00000000..2c85ac25
--- /dev/null
+++ b/watchdog/pi_zero/watchdog.py
@@ -0,0 +1,276 @@
+#!/usr/bin/env python3
+"""
+Watchdog script for Pyro Engine hardware.
+
+It checks the main Pi health endpoint and pings camera IPs, tracking failures
+and power-cycling relays after repeated failures with cooldown/daily limits.
+
+Cron setup (every 10 minutes):
+  1) Edit crontab:  crontab -e
+  2) Add the line:
+     */10 * * * * /usr/bin/python3 /home/pi/pyro-engine/watchdog/pi_zero/watchdog.py >> /home/pi/watchdog.log 2>&1
+
+Adjust the path to match where this repo lives on the Pi.
+"""
+
+import datetime as dt
+import json
+import logging
+import subprocess
+import time
+from dataclasses import dataclass
+from pathlib import Path
+from urllib.request import Request, urlopen
+
+import RPi.GPIO as GPIO
+
+# ================= CONFIG =================
+
+RELAY_MAIN = 16
+RELAY_CAMS = 26
+
+MAIN_PI_IP = "192.168.1.99"
+MAIN_HEALTH_URL = f"http://{MAIN_PI_IP}:8081/health"
+
+CAM_IPS = ["192.168.1.11", "192.168.1.12"]
+INTERNET_IP = "1.1.1.1"
+
+PING_COUNT = 2
+TIMEOUT = 2  # seconds, used for ping and HTTP timeout
+
+MAX_FAILS = 3
+POWER_OFF_TIME = 15
+
+COOLDOWN_SECONDS = 30 * 60
+MAX_REBOOTS_PER_DAY = 3
+
+STATE_DIR = Path("/tmp")
+LOG_FILE = Path("/home/pi/watchdog.log")
+
+FAIL_MAIN_FILE = STATE_DIR / "fail_main"
+FAIL_INTERNET_FILE = STATE_DIR / "fail_internet"
+FAIL_CAM_FILES = {ip: STATE_DIR / f"fail_cam_{ip.split('.')[-1]}" for ip in CAM_IPS}
+
+MAIN_LAST_REBOOT_FILE = STATE_DIR / "last_reboot_main"
+CAMS_LAST_REBOOT_FILE = STATE_DIR / "last_reboot_cams"
+MAIN_DAILY_FILE = STATE_DIR / "daily_reboots_main"
+CAMS_DAILY_FILE = STATE_DIR / "daily_reboots_cams"
+
+# ================ LOGGING =================
+
+logging.basicConfig(
+    filename=str(LOG_FILE),
+    level=logging.INFO,
+    format="%(asctime)s | %(levelname)s | %(message)s",
+    datefmt="%Y-%m-%d %H:%M:%S",
+)
+
+# ================ GPIO ====================
+
+GPIO.setmode(GPIO.BCM)
+GPIO.setwarnings(False)
+
+GPIO.setup(RELAY_MAIN, GPIO.OUT, initial=GPIO.HIGH)
+GPIO.setup(RELAY_CAMS, GPIO.OUT, initial=GPIO.HIGH)
+
+# ================ IO HELPERS ==============
+
+def read_int(path: Path, default: int = 0) -> int:
+    try:
+        return int(path.read_text().strip())
+    except Exception:
+        return default
+
+
+def write_int(path: Path, value: int) -> None:
+    path.write_text(str(value))
+
+
+def read_text(path: Path, default: str = "") -> str:
+    try:
+        return path.read_text().strip()
+    except Exception:
+        return default
+
+
+def write_text(path: Path, value: str) -> None:
+    path.write_text(value)
+
+# ================ CHECKS ==================
+
+def ping_host(ip: str, count: int = PING_COUNT, timeout: int = TIMEOUT) -> bool:
+    try:
+        subprocess.check_output(
+            ["ping", "-c", str(count), "-W", str(timeout), ip],
+            stderr=subprocess.DEVNULL,
+        )
+        return True
+    except subprocess.CalledProcessError:
+        return False
+
+
+def http_health_ok(url: str) -> bool:
+    req = Request(url, method="GET", headers={"accept": "application/json"})
+    try:
+        with urlopen(req, timeout=TIMEOUT) as resp:
+            if resp.status != 200:
+                return False
+            body = resp.read().decode("utf-8", errors="replace")
+        data = json.loads(body)
+        return data.get("status") == "ok"
+    except Exception:
+        return False
+
+
+# ================ FAIL COUNTERS ===========
+
+def update_fail_counter(ok: bool, fail_file: Path, label: str, log_result: bool = True) -> int:
+    if ok:
+        if log_result:
+            logging.info("%s check OK", label)
+        write_int(fail_file, 0)
+        return 0
+
+    fails = read_int(fail_file, 0) + 1
+    if log_result:
+        logging.warning("%s check FAILED (%s/%s)", label, fails, MAX_FAILS)
+    write_int(fail_file, fails)
+    return fails
+
+# ================ REBOOT GUARD ============
+
+@dataclass(frozen=True)
+class RebootGuard:
+    cooldown_seconds: int
+    max_reboots_per_day: int
+
+    def _read_daily(self, daily_file: Path) -> tuple[str, int]:
+        today = dt.date.today().isoformat()
+        raw = read_text(daily_file, "")
+
+        if not raw:
+            return today, 0
+
+        parts = raw.split()
+        if len(parts) != 2:
+            return today, 0
+
+        day, count_s = parts[0], parts[1]
+        try:
+            count = int(count_s)
+        except Exception:
+            count = 0
+
+        if day != today:
+            return today, 0
+
+        return day, count
+
+    def can_reboot(self, now_ts: int, last_reboot_file: Path, daily_file: Path, label: str) -> bool:
+        last_ts = read_int(last_reboot_file, 0)
+        if last_ts and (now_ts - last_ts) < self.cooldown_seconds:
+            remaining = self.cooldown_seconds - (now_ts - last_ts)
+            logging.warning("%s: reboot blocked by cooldown, remaining_seconds=%s", label, remaining)
+            return False
+
+        day, count = self._read_daily(daily_file)
+        if count >= self.max_reboots_per_day:
+            logging.warning(
+                "%s: reboot blocked by daily limit, count=%s, limit=%s",
+                label,
+                count,
+                self.max_reboots_per_day,
+            )
+            write_text(daily_file, f"{day} {count}")
+            return False
+
+        return True
+
+    def record_reboot(self, now_ts: int, last_reboot_file: Path, daily_file: Path) -> None:
+        write_int(last_reboot_file, now_ts)
+
+        today = dt.date.today().isoformat()
+        day, count = self._read_daily(daily_file)
+        if day != today:
+            day, count = today, 0
+
+        count += 1
+        write_text(daily_file, f"{day} {count}")
+
+def power_cycle(relay_gpio: int, label: str, last_file: Path, daily_file: Path, guard: RebootGuard) -> None:
+    now_ts = int(time.time())
+
+    if not guard.can_reboot(now_ts, last_file, daily_file, label):
+        return
+
+    logging.warning("%s: power cycle triggered", label)
+    GPIO.output(relay_gpio, GPIO.LOW)
+    time.sleep(POWER_OFF_TIME)
+    GPIO.output(relay_gpio, GPIO.HIGH)
+    logging.info("%s: power restored", label)
+
+    guard.record_reboot(now_ts, last_file, daily_file)
+
+# ================= MAIN ===================
+
+def main() -> None:
+    guard = RebootGuard(
+        cooldown_seconds=COOLDOWN_SECONDS,
+        max_reboots_per_day=MAX_REBOOTS_PER_DAY,
+    )
+
+    reboot_12v = False
+
+    internet_ok = ping_host(INTERNET_IP, count=1, timeout=TIMEOUT)
+    internet_fails = update_fail_counter(internet_ok, FAIL_INTERNET_FILE, f"Internet {INTERNET_IP}")
+    if internet_fails >= MAX_FAILS:
+        reboot_12v = True
+
+    if internet_ok:
+        main_ok = http_health_ok(MAIN_HEALTH_URL)
+        main_fails = update_fail_counter(main_ok, FAIL_MAIN_FILE, "Main Pi health")
+    else:
+        logging.warning("Skipping Main Pi health check (internet_ok=%s)", internet_ok)
+        main_fails = 0
+
+    if main_fails >= MAX_FAILS:
+        power_cycle(RELAY_MAIN, "Main Pi", MAIN_LAST_REBOOT_FILE, MAIN_DAILY_FILE, guard)
+        write_int(FAIL_MAIN_FILE, 0)
+
+    cam_results: list[tuple[str, bool, int]] = []
+
+    for ip in CAM_IPS:
+        ok = ping_host(ip)
+
+        fails = update_fail_counter(ok, FAIL_CAM_FILES[ip], f"Camera {ip}", log_result=False)
+        cam_results.append((ip, ok, fails))
+        if fails >= MAX_FAILS:
+            reboot_12v = True
+
+    if cam_results:
+        parts = []
+        any_failed = False
+        for ip, ok, fails in cam_results:
+            if ok:
+                parts.append(f"{ip} OK")
+            else:
+                parts.append(f"{ip} FAILED ({fails}/{MAX_FAILS})")
+                any_failed = True
+
+        summary = "Cameras check: " + ", ".join(parts)
+        if any_failed:
+            logging.warning(summary)
+        else:
+            logging.info(summary)
+
+    if reboot_12v:
+        power_cycle(RELAY_CAMS, "Cameras / Router 12V", CAMS_LAST_REBOOT_FILE, CAMS_DAILY_FILE, guard)
+        for ip in CAM_IPS:
+            write_int(FAIL_CAM_FILES[ip], 0)
+        write_int(FAIL_INTERNET_FILE, 0)
+
+if __name__ == "__main__":
+    try:
+        main()
+    finally:
+        GPIO.cleanup()