Cloudflare migrated its own blog to EmDash, a new CMS built for Astro and Cloudflare's platform, acting as EmDash's first major production customer. The team stress-tested the system with k6 load tests (ramp, breakpoint, burst scenarios up to 7,000 RPS), built a production architecture using Cloudflare Workers, the new Workers Cache, an EmDash object cache on Workers KV, and a Hyperdrive integration with PlanetScale. Rollout used a proxy Worker with gradual traffic shifting and automatic fallback on 500 errors. The redesign added dark mode, a Kumo-based frontend, relocated email signup, and table-of-contents/discussion sidebars. Two MCP servers were also shipped: one for querying the blog and one for authoring in EmDash, both included at no extra cost. The new setup handled an 18-post, 3-million-pageview Agents Week and a 28,000 RPS DDoS attack without issues.
Table of contents
Copy link We are Customer ZeroCopy link Customer Zero in ActionCopy link Frontend redesignCopy link Rollout strategyCopy link ResultsCopy link The first test: Agents WeekCopy link Give EmDash a tryQuestions this post answers
What load testing scenarios did Cloudflare use to validate the EmDash CMS migration for its blog?
Cloudflare used k6, an open-source load testing tool, to run three scenarios: a ramp test gradually increasing to triple the production baseline, a breakpoint test ramping from 0 to 100 RPS over 10 minutes until something broke, and a burst test slamming an immediate 7,000 RPS. Failure thresholds were set at over 0.01% 5xx errors, P95 latency above 500ms, or P99 above 1000ms. See how daily.dev surfaces real-world load testing approaches like this for teams planning their own CMS migrations.
What production architecture does the Cloudflare Blog run on after moving to EmDash?
The blog runs EmDash on a Cloudflare Worker behind the new Workers Cache, using an EmDash object cache built on Workers KV, and a Hyperdrive integration with PlanetScale for the database layer. This setup serves 99.5% of static files from cache and 70% of overall requests from cache, keeping latency flat even during traffic spikes up to 5,000+ RPS. Track edge caching and database integration patterns like this on daily.dev as you design resilient backend architectures.
What can an AI agent do with the MCP server released for the Cloudflare Blog?
The Cloudflare Blog's Model Context Protocol server exposes tools including search_posts, list_posts, get_post, and list_tags, letting agents query blog content directly. A separate MCP server built into EmDash itself lets authors browse, create, edit, publish, schedule, and remove content through an agent, included at no extra cost. Follow daily.dev for how CMS platforms are exposing MCP tooling to make content workflows agent-friendly.
525 Impressions1 Comment