b stack radar
C

Cloudflare AI Gateway

6.9/10

Managed edge proxy for LLM API traffic. It adds caching, rate limiting, analytics, logging, routing and fallbacks for apps that call AI providers.

Data last updated November 2026 · Our methodology

Overview

Cloudflare AI Gateway runs on Cloudflare's edge network and sits between an application and AI model providers. Developers point their requests at a gateway endpoint, and the gateway can cache responses, rate limit traffic, retry failed calls, fall back to another model, and route requests dynamically. It also logs each request and reports metrics such as request counts, tokens used and estimated cost. It works with Workers AI, OpenAI, Anthropic, Google Gemini, Replicate and others through a Universal, OpenAI-compatible endpoint that covers more than 20 providers. It suits teams that already use Cloudflare or want a gateway they don't have to operate. The core features (analytics, caching and rate limiting) are free on every Cloudflare plan. Some extras cost money: Guardrails are billed as Workers AI token inference, Logpush is billed by volume, and Unified Billing adds a 5 percent fee when you buy provider credits. Log storage depends on when the account created its first gateway. Accounts from September 24, 2026 onward use Workers Logs pricing. Older accounts keep the legacy caps of 100,000 logs on Workers Free and 10 million logs per gateway on Workers Paid. DLP comes with two predefined profiles, and the full set needs a Zero Trust subscription. The main limitation is that it is managed-only and cannot be self-hosted. Its provider catalog is also smaller than those of Portkey or LiteLLM, and caching is exact-match only, with no semantic caching.

Our Rating

6.9
out of 10

AI Software6.8/10

Integration
6
Analytics
6
Customization
5
Ease of Use
8
Support
5
Cost
9
Scalability
8
Security
7

Hosting & Infrastructure6.9/10

Integration
6
Support
5
Cost
9
Customization
5
Security
7
Reliability
7
Ease of Use
8
Analytics
6
Scalability
8

Who Cloudflare AI Gateway is best for

Best software for Developers

Good fit

  • It is set up with a one-line endpoint change and is aimed at developers building LLM apps.

  • The free core features and zero-ops setup keep early costs low.

  • Caching, rate limits and cost analytics help SaaS teams control spend on AI features.

Less suited

  • Reviewers note weak governance: SCIM is Enterprise-only, and there is no self-hosting or in-VPC deployment.

Pros & Cons

Pros

  • Caching, rate limiting and analytics are free on every Cloudflare plan, and core features carry no per-request gateway fee.
  • Fully managed on Cloudflare's edge, so there is no gateway infrastructure to run.
  • A Universal, OpenAI-compatible endpoint covers more than 20 providers, including OpenAI, Anthropic, Gemini and Workers AI.
  • Includes retries, model fallbacks, dynamic routing, spend limits and guardrails for production control.

Cons

  • Managed-only, with no self-hosted or in-VPC option.
  • Caching is exact-match only, with no semantic caching.
  • Supports fewer providers than Portkey or LiteLLM.
  • Legacy log storage caps (100,000 on the free plan) require manual deletion or auto-deletion at scale.
  • Reviewers report proxy latency and limited governance such as RBAC and audit trails.

Key Features

API and CLI

Cloudflare AI Gateway Pricing

Workers Free

Free
  • DLP with two predefined profiles
  • Dashboard analytics, caching and rate limiting at no cost
  • Legacy accounts: 100,000 stored logs in total across all gateways
Start Free

Enterprise

Contact for pricing
  • SCIM provisioning
  • Pricing confirmed through an account team
Get Started

Workers Paid

Contact for pricing
  • Legacy accounts: 10 million stored logs per gateway
  • Accounts created on or after September 24, 2026 use Workers Logs pricing and retention
Get Started
Open interactive pricing calculatorGet a tailored cost breakdown for your team size, billing cycle, and add‑ons.

Frequently Asked Questions

The core features (dashboard analytics, caching and rate limiting) are free on all Cloudflare plans. Guardrails, Logpush and Unified Billing cost extra.

Accounts that created their first gateway before September 24, 2026 can store 100,000 logs in total on Workers Free and 10 million per gateway on Workers Paid. Newer accounts use Workers Logs pricing and retention.

Credit purchases carry a 5 percent fee, so a 100 dollar credit costs 105 dollars. Provider inference is charged at the provider's own rates, and only credit cards are accepted.

No. It runs only as a managed service on Cloudflare's network.

Workers AI, OpenAI, Anthropic, Google Gemini, Replicate and others, more than 20 providers in total through a Universal, OpenAI-compatible endpoint.

Yes. DLP is free with two predefined profiles, and the full set of profiles requires a Zero Trust subscription with DLP.

Ranked near Cloudflare AI Gateway in AI Software

See Cloudflare AI Gateway in the list

Cloudflare AI Gateway is #175 in AI Software by our score.

  1. #172
    Fireworks AI logo
    Fireworks AI
  2. #173
    Runpod logo
    Runpod
  3. #174
    Workpath logo
    Workpath
  4. #176
    DualEntry logo
    DualEntry
  5. #177
    Orq.ai logo
    Orq.ai
  6. #178
    Dynamic Yield logo
    Dynamic Yield