Lanshu: an AI-assisted Cloudflare billing risk checklist
Lanshu (@LufzzLiz) shares a six-step prompt for checking account ownership, usage, runaway consumption and budget alerts. Replies add alarm logging, email delivery checks and separate notifications.
- Products
- WorkersDurable ObjectsD1R2KVQueuesContainersWorkers AI
- Source
- Author experience𝕏
- Author
- Lanshu (@LufzzLiz)
Key points
- The author asks an authorized AI assistant to confirm account and project ownership from domain bindings, deployment records and repository configuration rather than project names.
- The prompt inventories enabled products and checks current-cycle costs plus 7-day and 30-day usage trends, recording billing units.
- It checks alarm rescheduling, loops and retries without exit conditions, repeated database operations, sudden usage growth, public endpoints and credential abuse. Code inspection depends on available access.
- The author asks the assistant to consult current official documentation and distinguish alerts, free-tier hard limits, per-request resource limits and spending caps, including reporting delays.
- The suggested prompt requests $1, $5 and $10 alert thresholds using the existing billing email, preserving effective alerts and reading configuration back after saving. It asks about other channels and plan requirements; these are requested settings, not a report that all channels are supported.
- The prompt limits changes to budget alerts. Plan upgrades, deployments, service shutdowns, resource deletion, payment changes and credential revocation require separate confirmation, and secrets must not appear in the output.
Original sources
Original post
Read on 𝕏岚叔
@LufzzLiz · Original 𝕏 post
深表同情,我也是深度使用cloudflare的用户,让agent做了一轮深度检查,账单告警配置,但是好像免费用户只有邮箱告警,看来还需要联动personal AI agent,帮我们整理邮件。然后信用卡做消费限额。 附cloudflare 检查提示词: 这段可以直接复制给能访问 Cloudflare 控制台或 API 的 AI 助手:… https://t.co/XkJPdVk2t6 pic.twitter.com/rMUCer2a1i
Reply from LIght JUNction (@LIghtJUNction_x): log alarmInfo.isRetry/retryCount and the time and reason for setAlarm; distinguish platform retries from application rescheduling, including catch/finally. Budget alerts notify rather than cap usage.
Read on 𝕏LIght JUNction
@LIghtJUNction_x · Original 𝕏 post
可以给检查提示词再补一项:把 alarmInfo 的 isRetry/retryCount,以及每次 setAlarm 的时间和原因打日志。 Cloudflare 自动重试最多 6 次,但代码每轮重新 setAlarm 可以一直续下去。先区分平台重试和代码自己重新排期,尤其检查 catch/finally。 预算告警只通知,不会暂停或封顶用量。…
Reply from Zaji (@lijd426): test email delivery after configuring alerts, including whether messages land in spam.
Read on 𝕏杂记
@lijd426 · Original 𝕏 post
告警配好后也验一次能不能收到,别等出事才发现邮件进了垃圾箱
Reply from Xinan (@Thomas3cnl): send billing alerts separately rather than delaying them until the daily email digest.
Read on 𝕏辛安
@Thomas3cnl · Original 𝕏 post
这类账单告警最好别进每日邮件摘要。让 agent 收到后单独推送,免得凌晨的异常到早上才被看到。
More to read
Official docsAccount-wide usage-based products
Budget alerts
Cloudflare documents account-wide spend alerts and their notification limits.
Workers
Sitemap polling: the author reports a $100 Cloudflare bill alongside AWS costs.
@enzyme_dev says a refactor made Workers fetch a sitemap every five seconds. In an October 6 reply, he reports a $100 Cloudflare bill and an AWS bill above $800 for the previous week.
Author experienceWorkers +4
Cloudflare billing pitfalls thread
The author organizes Cloudflare billing risks by product and shares experience with budget alert emails.