# Cloudflare Is Quietly Building the Agent Cloud **Source doc**: [[3. Resources/Readwise/Articles/Cloudflare Is Quietly Building the Agent Cloud]] > [!question] Lingering Questions > - The central move is inferring corporate *intent* from a release cadence — plausible, but unfalsifiable from outside the company. What would actual evidence of "coordinated strategy" vs. "convergent but separate feature bets" look like? > - Are Cloudflare's MCP servers and Web Bot Auth implementations genuinely portable, or is this soft lock-in dressed as an open standard? > - The author teaches a Cloudflare-for-WordPress course — how much does that interest shape the framing? > [!tip] Article Ideas > - Benchmark the "one fabric = lower latency" claim: a Cloudflare-only agent stack vs. a cross-cloud one, on a real multi-step task. > [!link] Cross Reference > - [[AI & the Future of Knowledge Work MOC]] — agents-as-distributed-systems is a recurring reframe worth tracking across sources. > - [[3. Resources/Permanent Notes/AI Editing Tools Earn Trust By Re-Rendering The Real Platform]] — a smaller-scale instance of the same "verify the actual resulting state" discipline this article argues infrastructure should provide agents at scale. > [!quote] Vocabulary > - **Durable execution** — multi-step logic that survives partial failure and resumes correctly, rather than restarting or silently corrupting state. > - **Web Bot Auth** — Cloudflare's cryptographic-identity scheme for signed, verifiable automated agents. > - **AutoRAG** — a managed crawl/embed/store/recrawl pipeline that keeps a retrieval index from "rotting" out of date. ## Author's thesis Remkus de Vries argues that Cloudflare's "Markdown for Agents" release, read alongside a year of prior launches (Workflows, Durable Objects, AI Gateway, Workers AI, AutoRAG, MCP servers/Portals, Firewall for AI, Web Bot Auth), is not a scatter of unrelated features but a coordinated infrastructure stack for AI agents — and that competing at the infrastructure layer compounds in a way competing at the rapidly-commoditizing model layer does not. ## Key arguments (≤3) 1. **An agent is a distributed system wearing an LLM as its brain** — the model proposes the next action, but the surrounding system must execute it safely, reliably, and repeatedly, which is an infrastructure problem, not a prompting problem. 2. **Each infrastructure layer maps to a real distributed-systems failure mode agents hit**: long-running logic that must survive partial failure (Workflows), concurrent tool calls corrupting shared state (Durable Objects), retrieval indexes that rot (AutoRAG), ungoverned tool access (MCP Portals), unverifiable agent actions (Web Bot Auth). 3. **Infrastructure compounds while the model layer commoditizes** — most vendors are fighting where value is eroding fastest; Cloudflare is building where advantages accumulate. ## Claim atoms ### Agents are distributed systems with an LLM brain > [!quote] ID:rw_cf_agent_1 > "An agent is not a 'smart chatbot.' It is a distributed system whose decision-making component happens to be an LLM." Where/why: "LLM Brain" section, near the top of the argument — the load-bearing reframe everything downstream is justified by. ### Durable execution separates demos from systems > [!quote] ID:rw_cf_agent_2 > "Durable execution is the line between demos and systems." Where/why: "Durable execution" section — multi-step agent logic that must survive partial failure; bundling this into the platform beats assembling queues/cron/retries by hand. ### Cryptographic agent identity precedes an agent web > [!quote] ID:rw_cf_agent_3 > "Cryptographic identity for agents is not a feature. It is a prerequisite for an agent-driven web." Where/why: "Trust is the long game" section — once agents transact, receiving systems must distinguish authorized delegation from arbitrary automation. ### Infrastructure compounds while the model layer commoditizes > [!quote] ID:rw_cf_agent_4 > "Most companies are competing at the model layer, which is rapidly commoditizing. Cloudflare is competing at the infrastructure layer, which compounds." Where/why: closing section — the strategic claim of the essay; useful framing but asserted rather than evidenced. ## Permanent candidates - **Model an agent as a durable workflow, not a chatbot** — promoted below. The clearest, most generalizable claim: agent reliability is a distributed-systems property, not a prompting property. - The "infrastructure compounds, models commoditize" claim is compelling but is an assertion about competitive dynamics, not evidence — held back from Permanent status pending some independent read on Cloudflare's actual market position. ## Processing checklist - [ ] Check whether Cloudflare's MCP / Web Bot Auth implementations are genuinely portable or soft lock-in - [ ] Tie the Web Bot Auth framing into any courtneyr.dev agent-readiness notes - [x] Connect into the relevant MOC — [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] ## Links - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — agents-as-distributed-systems reframes what "AI-native" infrastructure needs to provide. - [[3. Resources/MOCs/AI Agent Safety MOC]] — Web Bot Auth's cryptographic-identity argument is an agent-safety infrastructure question at the network layer.