Cloudflare AI Gateway: what it covers for model traffic
Blog post from Barndoor
Cloudflare AI Gateway is a proxy for application-to-model traffic that can be adopted through a base-URL change and provides OpenAI-compatible routing, caching, rate limiting, fallbacks, spend analytics and limits, request logging, DLP, guardrails, and key-management options, with several features still in beta. Core analytics, caching, and rate limiting are free, while services such as expanded DLP, log storage, Logpush, Guardrails, and Unified Billing can add costs. Identity and spending attribution rely on client-supplied metadata unless the gateway is placed behind Cloudflare Access, which can add a verified user identifier but requires Zero Trust and a custom domain. The text argues that Cloudflare separates model traffic from MCP tool traffic, which is managed through Cloudflare One MCP server portals with separate policies and logs, creating limitations for organizations seeking unified governance across users, agents, models, and tools. It identifies gaps in documented per-agent identities, immutable attribution, fine-grained tool and action permissions, approval workflows for risky writes, and a single SIEM-ready audit trail. It concludes that Cloudflare AI Gateway is well suited to teams focused on application-to-model reliability and cost management, while presenting Barndoor as an enterprise-focused alternative intended to unify policy enforcement, identity, approvals, data protection, budgets, and auditing across model and MCP tool calls.
| Trend | Post Mentions | Total Month Mentions | Posts | Companies | MoM |
|---|---|---|---|---|---|
| MCP | 21 | 11,940 | 1,236 | 235 | +13% |
| Zero Trust | 7 | 140 | 52 | 28 | +24% |
| LLM | 5 | 6,578 | 1,243 | 236 | +9% |
| Platform Engineering | 3 | 2,014 | 399 | 102 | +19% |
| Observability | 1 | 3,086 | 746 | 186 | +11% |
| Secrets Management | 1 | 3,098 | 623 | 155 | +51% |
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.