CloudFlare Just Dropped Their AI Engineering Stack

Cloudflare’s Internal AI Engineering StackCloudflare just dropped a solid post on the AI platform they built for themselves using the same tools they ship to customers.

Key wins: 93% adoption across R&D, weekly merge requests jumped from ~5,600 to over 8,700, and they’re running hundreds of millions of AI requests per month — all with strong controls on cost, security, and quality.

What matters for us :

A clean internal AI gateway for safe, observable access to models
AGENTS.md files in repos to give agents real context and reduce hallucinations
An automated AI Code Reviewer baked into CI that checks quality, security, and their engineering standards
Smart architecture decisions that keep token costs and context under control

It’s a practical blueprint for scaling agentic AI without chaos.

If you’re working on our AI tooling, code quality, or internal efficiency, this one’s worth 10 minutes.→ Read it here: https://blog.cloudflare.com/internal-ai-engineering-stack/

Thoughts? Which piece should we try first — AGENTS.md the gateway, or the reviewer?

15 comments