claude-watchdog
由 chapati23 提供
DevOps & Cloud简介
Monitor the Claude API for outages and latency spikes with rich Telegram alerts.
描述
该技能的功能
Monitors the Anthropic API for outages and latency spikes via cron-based polling. Sends structured Telegram alerts for status incidents, latency spikes, and recoveries, tagging whether an incident affects your specific model. Status checks cost zero tokens; latency probes measure real end-to-end response time through the local OpenClaw gateway.
Status checks poll the official Anthropic status page directly at zero token cost, and the latency probe measures your actual gateway response path rather than relying on third-party synthetic monitoring.
何时使用
- Getting notified on Telegram when the Claude API goes down
- Detecting latency spikes before users report slow responses
- Filtering out incidents that affect other models but not yours
- Receiving an automatic all-clear message when the API recovers
- Distinguishing whether slowdowns are Anthropic-side or local
示例工作流
User asks: set up Claude API watchdog alerts for my Sonnet usage
- 1
Runs setup.sh interactively, prompting for Telegram bot token, chat ID, and gateway credentials
- 2
Writes config to ~/.openclaw/skills/claude-watchdog/claude-watchdog.env with MONITOR_MODEL=sonnet
- 3
Installs two cron jobs running every 15 minutes: status-check.py and latency-probe.py
- 4
Runs an initial status check and latency probe to verify connectivity and establish a baseline
- 5
Confirms setup by sending a test alert to the configured Telegram chat
Cron jobs installed; Telegram receives incident alerts tagged to your model and latency warnings with severity levels automatically going forward
前置要求
- Telegram bot token from @BotFather (requires creating a Telegram bot)
- Telegram chat ID for the channel or group where alerts will be delivered
安装
clawhub install chapati23/claude-watchdog运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。