Cloudflare AI Gateway
LLM proxy with free caching, rate limits and spend caps across providers
Overview
Cloudflare AI Gateway sits between your app and providers such as OpenAI, Anthropic, Google and Workers AI, adding analytics, exact-match caching, rate limits, spend limits, retries and fallback behind one endpoint. Core features are free on all plans; logs are priced by Workers plan, and Unified Billing adds a 5% fee on credits.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 12 Oct 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Dashboard analytics
- Caching
- Rate limiting
- DLP: financial and national-ID profiles
- Workers Free: 200,000 events a day, 3-day retention
- Workers Paid: 20M a month included, +$0.60 per extra million, 7-day retention
- Cloudflare Observability pricing from 1 Dec 2026
- Workers Free: 100,000 logs total across all gateways
- Workers Paid: 10,000,000 logs per gateway
- 10 million requests a month included
- +$0.05 per extra million
- Billed as token-based llama-guard-3-8b inference
- 5% fee on credits purchased
- Provider inference passed through with no markup
Use cases
What you can produce with Cloudflare AI Gateway
- One endpoint and REST API for OpenAI, Anthropic, Google, Workers AI and other providers
- Exact-match response caching, request retries and model fallback
- Token, cost and latency analytics, plus a per-user spend view in User Insights
- Spend limits by model, provider or metadata, and per-gateway rate limiting
- Free DLP scanning with financial and national-ID profiles
- One Cloudflare credit balance for third-party and Workers AI models through Unified Billing
AI Gateway puts a free control layer in front of LLM API calls: one endpoint for many providers, with token and cost analytics, exact-match caching, rate limiting, spend limits, retries and model fallback. Inference is passed through at provider rates, and Unified Billing lets one Cloudflare credit balance pay for third-party models for a 5% fee on credits.
The costs sit in logging. Accounts that create their first gateway on or after 24 September 2026 follow Workers Logs pricing: 200,000 events a day with three-day retention free, or 20 million a month included on Workers Paid. Caching is exact-match only, and the docs list no MCP feature. A sensible default for teams already on Cloudflare.
What we found
Cloudflare AI Gateway is a proxy that sits between an application and its model providers. Cloudflare opened it as a beta in September 2023 and made it generally available in May 2024. Point existing OpenAI, Anthropic, Google or Workers AI calls at a gateway and you get request and token analytics, exact-match response caching, rate limiting, retries and model fallback, with the core features free on every plan.
Most of the 2026 changes are about money. Spend limits arrived in June, capping dollar spend by model, provider or custom metadata rather than by request count. Unified Billing lets one prepaid Cloudflare credit balance pay for third-party models and, since August, Workers AI; Cloudflare adds a 5% fee to each credit purchase but passes provider token prices through without markup. User Insights, added in August, attributes spend to individual users, flags anomalous sessions and points out requests a cheaper model could handle.
Logging is where the bill appears, and it changed on 24 September 2026. Accounts that create their first gateway on or after that date follow Workers Logs pricing: 200,000 log events a day with three-day retention on Workers Free, or 20 million a month plus $0.60 per extra million with seven-day retention on Workers Paid, which starts at $5 a month. Older accounts keep the legacy caps of 100,000 stored logs in total on Free and 10 million per gateway on Paid; an earlier version of this page gave both figures wrongly. From 1 December 2026 those logs move to Cloudflare's Observability pricing.
The gaps are documented: caching is exact-match only, with semantic caching described as planned, and the AI Gateway docs list no MCP feature. Free DLP covers two predefined profiles, financial information and national identifier numbers.
For a team already on Cloudflare it is a cheap way to add cost visibility and guardrails; it is not a reason to switch platforms.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
About this listing
This entry was compiled from publicly available data including Cloudflare AI Gateway's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Cloudflare AI Gateway unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Cloudflare AI Gateway directly →
Spotted something out of date? Suggest an update →
Related reading
More in Security & Safety
