Cloudflare AI Gateway

Spec sheet

What it does

Cloudflare AI Gateway is a managed control plane that routes AI application traffic through one gateway for visibility, controls, billing and logs.Read 21 Oct 2026 · cloudflare.com

Headquarters

San Francisco, CARead 23 Sep 2026 · cloudflare.com

Founded

2009Read 23 Sep 2026 · cloudflare.com

Analytics

The dashboard reports requests, tokens, costs, errors, cache-hit rates and performance across providers.Read 21 Oct 2026 · developers.cloudflare.com

Authentication

Requests use Cloudflare API tokens, while upstream credentials can be supplied through Unified Billing, stored keys or request headers.Read 21 Oct 2026 · developers.cloudflare.com

Auto Router

Auto Router selects models based on task suitability and cost, with fallback when a model is unavailable.Read 21 Oct 2026 · developers.cloudflare.com

BYOK

Provider keys can be stored in Cloudflare's encrypted infrastructure, centrally rotated and used at request time.Read 21 Oct 2026 · developers.cloudflare.com

Caching

Responses can be served from Cloudflare's global cache, with configurable TTL and per-request controls; Cloudflare says latency can fall by up to 90% for repeated queries.Read 21 Oct 2026 · developers.cloudflare.com

Company founded and headquarters

Cloudflare was founded in 2009 and is headquartered in San Francisco, California.Read 21 Oct 2026 · linkedin.com

Company legal name

The maker is Cloudflare, Inc., a public company listed on NYSE under NET.Read 21 Oct 2026 · crunchbase.com

Company size

LinkedIn lists Cloudflare's company size as 1,001–5,000 employees.Read 21 Oct 2026 · linkedin.com

Custom providers

An API lets customers create and manage HTTPS custom-provider endpoints, including internal or self-hosted model services.Read 21 Oct 2026 · developers.cloudflare.com

DLP

AI Gateway DLP scans prompts and responses for sensitive information and is presented as helping with GDPR, HIPAA and PCI DSS requirements.Read 21 Oct 2026 · developers.cloudflare.com

Documentation depth

The documentation includes setup, provider guides, REST reference, bindings, integrations, features, limits, pricing, troubleshooting and a product changelog.Read 21 Oct 2026 · developers.cloudflare.com

Dynamic routing

Visual routing flows support geographic or user segmentation, A/B traffic splits, content-aware rules and fallbacks.Read 21 Oct 2026 · developers.cloudflare.com

Guardrails

Guardrails inspect prompts and model responses for harmful content and can flag or block configured categories.Read 21 Oct 2026 · developers.cloudflare.com

Licence

The hosted AI Gateway service is not presented as an open-source project; an independent review notes there is no source project to self-host or patch.Read 21 Oct 2026 · llmtools.cc

Limit: free-tier constraints

Independent analysis warns that log quotas and other hard limits can constrain the free offering and may contribute to unexpected bills.Read 21 Oct 2026 · scored.tools

Limit: observability

Gartner reviewers report that deeper debugging and observability can be limited and documentation may be fragmented.Read 21 Oct 2026 · gartner.com

Limit: portability

Independent reviewers identify cloud-only deployment and lack of self-hosting as a limitation for portability-sensitive teams.Read 21 Oct 2026 · llmtools.cc

Log limits

Legacy logging allows 100,000 stored logs on Workers Free and 10 million per gateway on Workers Paid; Logpush allows four jobs per account.Read 21 Oct 2026 · developers.cloudflare.com

Notable news

Cloudflare announced identity-aware AI Gateway and User Insights on August 5, 2026, adding verified-user attribution, spend controls and anomaly signals.Read 21 Oct 2026 · cloudflare.net

Praise: cost

An independent review highlights free core features and no percentage markup on provider tokens as major advantages.Read 21 Oct 2026 · llmtools.cc

Praise: integration

Gartner reviewers of Cloudflare API Gateway repeatedly praise straightforward Cloudflare ecosystem integration, edge performance and built-in security controls.Read 21 Oct 2026 · gartner.com

Provider integrations

Native provider routes include Workers AI, Amazon Bedrock, Anthropic, Azure OpenAI, Baseten, Cartesia, Cerebras, Cohere, Deepgram, DeepSeek, ElevenLabs, Fal, Google AI Studio, Google Vertex, Groq, Hugging Face, Ideogram, Mistral, OpenAI, OpenRouter, Parallel, Perplexity, Replicate and xAI.Read 21 Oct 2026 · developers.cloudflare.com

Rate limiting

Rate limits can use fixed or sliding windows and apply per gateway or request to control scaling and abuse.Read 21 Oct 2026 · developers.cloudflare.com

Rating on Gartner Peer Insights

4.3 out of 5 from 67 ratings for Cloudflare API Gateway.Read 21 Oct 2026 · gartner.com

Rating on scored.tools

7.0 out of 10 from an independent editorial score.Read 21 Oct 2026 · scored.tools

Rating on Techjockey

No rating; the page says it is the first review and shows no submitted reviews.Read 21 Oct 2026 · techjockey.com

REST API

The REST interface provides /ai/run, OpenAI-compatible chat and responses endpoints, and an Anthropic-compatible messages endpoint.Read 21 Oct 2026 · developers.cloudflare.com

SDK integration

Cloudflare documents integration with the Vercel AI SDK through createAI from the Agents SDK; the older ai-gateway-provider package is deprecated.Read 21 Oct 2026 · developers.cloudflare.com

Security compliance

Cloudflare publishes SOC 2 materials and states controls aligned to the AICPA Trust Services Criteria.Read 21 Oct 2026 · cloudflare.com

Spend limits

Budget rules can be scoped by model, provider or custom metadata and can block requests after a time-window budget is exceeded.Read 21 Oct 2026 · developers.cloudflare.com

System requirements

The Workers setup guide requires a Cloudflare account and Node.js, with Wrangler used for local development and deployment.Read 21 Oct 2026 · developers.cloudflare.com

Unified billing

Unified Billing lets Cloudflare-managed credentials pay supported provider usage from prepaid credits; Cloudflare documents a 5% credit-purchase fee.Read 21 Oct 2026 · developers.cloudflare.com

Usage limits

Documented limits include 25 MB cacheable requests, one-month cache TTL, 10 free gateways per account, 20 paid gateways and 200 Unified Billing requests per gateway per 60 seconds.Read 21 Oct 2026 · developers.cloudflare.com

Workers binding

Workers can call models and gateway features through the env.AI binding configured in Wrangler.Read 21 Oct 2026 · developers.cloudflare.com

Plans and prices

Every Cloudflare AI Gateway plan and price →

Where Specifiction places it

• LLM Gateway Software#5 of 348.7
• LLM gateways#18 of 337.9