What changed
Cloudflare has introduced Pay Per Use, a beta that lets AI companies make offers for specific uses of publisher content. A publisher can accept or decline each offer, the buyer reports each use, and Cloudflare handles billing and monthly payouts. The model is deliberately different from charging for every crawl: the proposed payment follows a citation, recommendation, report, or other downstream use of the content.
The same announcement introduces Monetization Gateway, also in beta, for request-level payments on APIs, Model Context Protocol (MCP) tools, sites, and datasets. It uses the open x402 protocol so a client can receive an HTTP 402 Payment Required, sign an authorization, retry the request, and get the resource after the gateway verifies and settles the payment.
How the product works
For Pay Per Use, the AI buyer defines what counts as a paid use and sets a price. Publishers review the offer in Cloudflare’s dashboard and can stop participating later. Approved buyers identify their crawling activity with Verified Bots, then report each use as JSON containing a timestamp, source URL, and event ID. Cloudflare aggregates the reports, charges the buyer, and pays the publisher.
Monetization Gateway moves the same ideas closer to an API boundary. A seller can match requests by URL, headers, query parameters, or caller attributes, then choose fixed or variable pricing and whether to charge everyone or only verified bots. The getting-started guide currently documents a minimum price of $0.001. At the origin, the seller still has to validate the signed PAYMENT-CONTEXT before returning paid content.
Why it matters
The interesting shift is not simply that AI traffic can be charged. Cloudflare is productizing the missing middle between “let agents crawl everything” and “block every bot”: identity, permission, metering, settlement, and analytics in one control plane. For publishers, that creates a path to say yes on explicit terms instead of choosing between licensing negotiations and unrestricted access. For AI builders, it turns agent access into a commercial interface that can be priced and audited like any other product surface.
That also changes what a useful web endpoint looks like. A site may offer free pages for discovery, then charge when an AI search result cites a page or when a shopping agent uses a review. An API or MCP tool can expose a paid operation without inventing a custom checkout flow. The Monetization Gateway documentation shows how Cloudflare applies x402 payment requirements inside the HTTP exchange and supports both fixed and usage-dependent settlement.
There are important limits: Pay Per Use is a beta with buyer-reported usage, Monetization Gateway is in closed beta, and the current eligibility documentation restricts buyers and sellers to the United States. The commercial contract and reporting trust model still matter; the protocol does not make a questionable use legitimate by itself.
Builder takeaway
If you are building an agent-facing product, design the economic boundary as deliberately as the technical one:
- Separate free discovery from paid execution or downstream use.
- Define a paid use precisely and attach a stable event ID so it can be reconciled later.
- For APIs and MCP tools, prefer a signed request/response payment flow over bespoke billing glue when the ecosystem supports it.
- Keep authorization, policy, and usage records visible to the person or team that owns the content.
Cloudflare’s bet is that the agentic web needs a way to measure value after retrieval, not just a way to control the first request. If that pattern holds, “agent access” becomes a product capability with a price, a policy, and a receipt.