claude-watchdog

# Claude Watchdog 🐕

Monitor the Anthropic/Claude API for outages and latency spikes. Sends rich alerts to Telegram — no agent tokens consumed for status checks.

## What It Does

### Status Monitor (`status-check.py`)
- Polls `status.claude.com` every 15 minutes via cron
- Alerts with incident name, latest update text, per-component status
- Tags incidents as "(not our model)" if e.g. Haiku is affected but you use Sonnet
- Sends all-clear on recovery
- **Zero token cost**

### Latency Probe (`latency-probe.py`)
- Sends a minimal request through OpenClaw's local gateway every 15 minutes
- Measures real end-to-end latency to Anthropic API
- Maintains rolling baseline (median of last 20 samples)
- Alerts with 🟡/🟠/🔴 severity based on spike magnitude
- Sends all-clear when latency recovers
- **~$0.000001 per probe**

## Setup

Run the interactive setup script:

```bash
bash /path/to/skills/claude-watchdog/scripts/setup.sh
```

You'll need:
1. **Telegram Bot Token** — from [@BotFather](https://t.me/BotFather)
2. **Telegram Chat ID** — send a message to your bot, then check `https://api.telegram.org/bot<TOKEN>/getUpdates`
3. **OpenClaw Gateway Token** — run:
   ```bash
   python3 -c "from pathlib import Path; import json; print(json.load(open(Path.home() / '.openclaw/openclaw.json'))['gateway']['auth']['token'])"
   ```
4. **Gateway Port** — default `18789`

The setup script writes config, installs cron jobs, and runs an initial check.

To uninstall (removes cron jobs, optionally config/state):
```bash
bash /path/to/skills/claude-watchdog/scripts/setup.sh --uninstall
```

## Config

Stored in `~/.openclaw/skills/claude-watchdog/claude-watchdog.env`. To reconfigure, either re-run `setup.sh` or edit this file directly — changes take effect on the next cron run (within 15 minutes).

```
TELEGRAM_BOT_TOKEN=...
TELEGRAM_CHAT_ID=...
OPENCLAW_GATEWAY_TOKEN=...
OPENCLAW_GATEWAY_PORT=18789
MONITOR_MODEL=sonnet
PROBE_MODEL=openclaw
PROBE_AGENT_ID=main
```

| Variable | Default...