ACCORDING to Cloudflare, in the last 30 days 93% of its R&D organisation used AI coding tools powered by infrastructure built on its own platform. Eleven months ago the team began integrating AI into the engineering stack, forming a tiger team called iMARS (Internal MCP Agent/Server Rollout Squad) to build the MCP servers, access layer, and AI tooling for agents.
The internal metrics show 3,683 internal users actively using AI coding tools, 47.95 million AI requests, 295 teams using agentic AI tools, 20.18 million AI Gateway requests per month, 241.37 billion tokens routed through AI Gateway, and 51.83 billion tokens processed on Workers AI. The architecture relies on a proxy Worker, Cloudflare Access, AI Gateway, and a discovery endpoint to configure everything automatically, while enabling per-user cost tracking and data retention controls.
As adoption nears its peak, the post notes a four-week rolling average rise in AI tooling usage and highlights Roadmap items like Code Mode at the portal layer and background agents to run in the cloud.