{"slug":"cloudflare-builds-an-internal-agentic-ai-engineering-stack-on-its-own-platform-reaching-93-r-d-adoption","url":"https://findausecase.com/use-cases/cloudflare-builds-an-internal-agentic-ai-engineering-stack-on-its-own-platform-reaching-93-r-d-adoption","title":"Cloudflare builds an internal agentic AI engineering stack on its own platform, reaching 93% R&D adoption","description":"Cloudflare describes an eleven-month internal rollout of AI coding tools routed through AI Gateway behind Cloudflare Access, with Workers AI inference, an MCP server portal of 13 servers exposing 182+ tools, AGENTS.md files generated for roughly 3,900 repositories, and an AI Code Reviewer covering all repos on its standard CI pipeline. The source reports 3,683 active users (60% of the company, 93% of R&D), 47.95 million AI requests, and a rise in weekly merge requests from about 5,600 to over 8,700 on a four-week average.","company":"Cloudflare","industry":"Technology & Software","aiCapabilities":["Agentic AI","Generative AI"],"deployment":"Unknown","problemStatement":"Cloudflare needed to integrate AI into its engineering stack, building the internal MCP servers, access layer, and AI tooling necessary for agents to be useful, and to rethink how standards are codified, how code gets reviewed, how engineers onboard and how changes propagate across thousands of repos.","solutionApproach":"An internal tiger team called iMARS built the stack on Cloudflare's own products: Cloudflare Access for authentication, AI Gateway for LLM routing, cost tracking and data retention controls, Workers AI for on-platform inference, and an MCP Server Portal aggregating 13 production MCP servers exposing 182+ tools. AGENTS.md files were generated for roughly 3,900 repositories, and an AI Code Reviewer runs on every merge request in the standard CI pipeline.","businessValue":"Cloudflare reports that in the last 30 days 3,683 internal users (60% company-wide, 93% across R&D) used AI coding tools, with 47.95 million AI requests. It also reports that the 4-week rolling average of merge requests climbed from ~5,600/week to over 8,700 as AI tooling adoption grew. These are company-reported figures.","evidence":{"band":"high"},"sourceUrl":"https://blog.cloudflare.com/internal-ai-engineering-stack/","dates":{"publishedAt":"2026-11-03T05:50:17.014Z","publishedAtSource":"pipeline","updatedAt":"2026-11-03T05:50:17.014Z"},"license":"Open for reading and citing with a link to https://findausecase.com/use-cases/cloudflare-builds-an-internal-agentic-ai-engineering-stack-on-its-own-platform-reaching-93-r-d-adoption. Bulk republication requires permission."}