Key facts
- Spend Management is opt-in and lets customers set spending thresholds so Cloudflare can pause activity automatically when limits are reached, avoiding unexpected bills from unplanned usage.
- Customers can enable controls through the dashboard, cf CLI, or REST API by defining a spending limit, specifying a notification email, and setting one or more alert thresholds.
- Cloudflare sends email notifications when evaluated spend reaches an alert threshold or the spending limit.
- When the spending limit is reached, a blocking policy pauses the product automatically and new invocations are rejected.
Cloudflare introduced Spend Management in Beta on Nov. 13, 2026, making it available to any pay-as-you-go customer, beginning with Workers and Durable Objects.
The opt-in tool lets customers set spending thresholds, and the system pauses activity automatically when spending reaches limits to avoid unexpected bills from unplanned usage. Through the dashboard, cf CLI, or REST API, customers define a spending limit, provide a notification e-mail address, and set one or more alert thresholds. Cloudflare sends e-mail notifications when evaluated spend reaches an alert threshold or the spending limit.
Once spending reaches the limit, the blocking policy pauses the product automatically and rejects new invocations. Enforcement halts only customer-code invocations that incur costs. In Workers, Cron Triggers and Queue consumer invocations stop, though static assets keep running. Customers restore service by raising the limit above accumulated spend, removing the blocking policy, or waiting for the block to expire at the start of the next billing cycle.
The dashboard displays current billing cycle costs for supported products, updating in near real time within 10 minutes, and keeps an audit log of limit changes and cap events. Billing cycles start on the paid subscription date, not when a customer deploys a Worker or creates a cap; a subscription beginning October 10 runs to November 10. Paid plans had access to usage notifications and budget alerts that only inform customers rather than controlling spending, and spend controls ranked among the most requested features from the more than 10 million developers on its platform.
AI Gateway spend limits, which support per-user and per-model budgets for Workers AI and other AI providers on gateway-routed requests, work alongside the new tool. The company plans to extend metering and spending caps to more products and primitives soon. Cloudflare also aims to close the gap between measurement and enforcement from under 10 minutes to under a minute over time. Planned application-based and agent-based budgets will let developers tag calls, attribute spending, and pause an application or agentic task across products without interrupting unrelated applications.
The Rating Engine calculates accumulated spend from usage records, subscription details and pricing rules, applying included allowances before it prices excess usage. Usage and subscription updates arrive as independent streams, and usage received before subscription details are available stays pending until pricing data arrives. Rating processing stays out of the request path; caps propagate asynchronously through Quicksilver to runtime nodes during periodic refresh. A two-stage metering pipeline aggregates raw invocation measurements into one-minute buckets and merges them into five-minute windows.
On Nov. 13, 2026, Cloudflare said the Usage Engine converts aggregated measurements into standardized usage records carrying account, usage window, metric, quantity and dimensions, and maps product meters to supported metrics. The Usage Engine partitions processing by account, with each partition holding local state and a durable write-ahead logging changelog on disk. Partitioning by account reduces cross-partition coordination, supports consistent pricing and parallel processing, and lets replacement processors rebuild and resume after failures.
On Nov. 13, 2026, Cloudflare said a product catalog database defines usage metrics with allowances, rates and rounding rules, and the engine prices accumulated usage rather than charging each request independently. One million CPU milliseconds does not automatically trigger an extra charge because the engine applies the included CPU allowance first and prices only the excess. After spend is calculated, policy evaluation checks warning or blocking thresholds, and reporting shows estimated usage charges through APIs and Cloudflare Email Service alerts.
Caps move asynchronously through the runtime configuration service and Quicksilver, an internal key-value store for fast global replication, and runtimes collect them on a periodic schedule; Cloudflare built the system so other products can onboard and expects the catalog to grow. Requests never wait on the Rating Engine because enforcement reuses Cloudflare’s existing runtime account-limit architecture. The Nov. 13, 2026, API example sets a 10000-cent threshold with action usage_block and notification e-mail me@example.com, while cf CLI billing spend-policy commands include list, create, get, update, delete, audit list and activity list.
The Nov. 13, 2026, dashboard example shows Workers triggering two alerts before reaching the hard limit. An overview panel on the dashboard lists which products have thresholds and limits configured.

Join the discussion