Cloudflare AI Gateway
Cloudflare AI Gateway is a proxy in front of LLM APIs. Rechecked 2026-08-17: developers.cloudflare.com/ai-gateway/reference/pricing says core features are free on all plans (dashboard analytics, caching, rate limiting) if you have a Cloudflare account. Persistent logs are on all plans, with Workers Free: 100,000 logs total and Workers Paid: 10,000,000 logs per gateway. Logpush is Workers Paid only. The paid-plan request row on that page is 10 million / month, then +$0.05 / million. A 5% fee applies to credits purchased through the gateway's credit flow. DLP scanning is free; accounts without Zero Trust get two predefined DLP profiles. github.com/cloudflare/ai had 1,143 stars, MIT.
Compare OpenRouter if you wanted a multi-provider wallet with a 5.5% Stripe fee, and Hugging Face if you wanted Hub-hosted inference instead of a proxy.
Key Features
- One line of code: Point your OpenAI-compatible client at the gateway. Recheck the current provider list in docs.
- Cache, rate limit, analytics: These are the free core. They do not make provider tokens free.
- Logs: Free accounts share a 100k log cap. Paid gets 10M per gateway.
- Not a model vendor: You still pay OpenAI / Anthropic / Workers AI separately.
Use Cases
- Teams already on Cloudflare who want one dashboard for token cost and cache hit rate.
- People who will stay on Workers Free and accept the 100k log cap.
- People comparing AI Gateway vs OpenRouter: one is a Cloudflare proxy; the other is a credit wallet.
Limitation: The old "hides source IP / bypasses access limits" line is not first-party pricing copy. We did not keep it. Enterprise details are "ask your account team."
Pricing
AI Gateway pricing on 2026-08-17.
| Path | Price | Notes from that page |
|---|---|---|
| Core gateway | $0 | Analytics, caching, rate limiting. Needs a Cloudflare account. |
| Workers Free logs | $0 | 100,000 logs total across gateways. |
| Workers Paid logs | Included on Paid | 10,000,000 logs per gateway. Logpush on Paid. |
| Paid extra requests | +$0.05 / million | After 10 million / month on the Paid row. |
| Gateway credits | +5% | Fee on credits purchased through that flow. |
If a leftover blog still says the gateway is an unmetered universal LLM, treat this table as the source.
Getting Started
- Read AI Gateway and create one gateway in the dashboard.
- Change only the base URL on an existing OpenAI-compatible client.
- Recheck Workers Free vs Paid before you turn on Logpush.
- Compare OpenRouter if you needed a provider wallet, not a Cloudflare proxy.
Frequently Asked Questions
Is AI Gateway free?
Core features are. Logs and extra requests follow your Workers plan.
Does it replace OpenRouter?
No. It proxies. You still need provider keys or Workers AI.
Same as Workers AI?
Workers AI is a model runtime. AI Gateway sits in front of many providers, including Workers AI.
Alternatives
- OpenRouter: Multi-provider credits.
- Hugging Face: Hub plus hosted inference.
Tips
- Quote free core plus 100k / 10M log caps.
- Do not paste "unmetered universal endpoint" into a deck.
- Recheck the $0.05 / million row before a high-QPS launch.
Comments
No comments yet. Be the first to comment!
Related Tools
OpenRouter
www.openrouter.ai
Unified LLM router. Rechecked: pay-as-you-go credits, 5.5% Stripe fee, BYOK free to $25k/month then 5%. Not a GPT-4o catalog page.
CheaperInference
cheaperinference.com
CheaperInference resells discounted AI inference through one OpenAI-compatible API, cutting model costs by up to 30% with usage-based billing, no contract.
SiliconFlow
www.siliconflow.cn
China inference API for open models. Official site siliconflow.cn. Qwen3-8B listed at ¥0.27/¥0.88 per 1M tokens. Skip unaudited 6M-user claims.
Related Insights
Skills + Hooks + Plugins: How Anthropic Redefined AI Coding Tool Extensibility
An in-depth analysis of Claude Code's trinity architecture of Skills, Hooks, and Plugins. Explore why this design is more advanced than GitHub Copilot and Cursor, and how it redefines AI coding tool extensibility through open standards.
After I Connected Obsidian to OpenClaw, It Started Helping Me Make Decisions
Once Obsidian stopped being just a place to store notes and started working with OpenClaw, it began helping me organize context, connect information, and improve real decisions.