# Cloudflare Master Knowledge Base — Complete Conversation Export **Generated:** July 23, 2026 > This file preserves every substantive user and assistant message in the Cloudflare research and publishing conversation through delivery of the master export. Chat-only `sandbox:` links are preserved as historical text and will not function on a public Cloudflare deployment. --- # Turn 1 · User **Category:** Research brief **Title:** Original deep-research request I want you to do an insanely deep, detailed analysis of every single possible function and capability of the CloudFlare.com platform. I'd like for you to do a summary of the history of CloudFlare: how the company started and how it's evolved. They claim on their website that 20% of the internet runs on CloudFlare. I want you to pull all of the news and the history: - the headlines - investment history - leadership - grows - leadership changes - the entire section on the evolution of the entire growth and leadership - number of employees - stock - patterns - private equity investments and ownership, etc. I want you to go research every possible page imaginable for CloudFlare to tell me every single bit of its feature sets, including the most current, every single element of what it can do, all of the capabilities and abilities within the platform. I want examples of all the different companies and products that host on CloudFlare. Provide links to those and if you don't know directly, just estimate what types of features it uses on CloudFlare so I can understand exactly how sophisticated products I can build on CloudFlare and list all of its functionality. Give examples of how those elements could be used. I am not a developer. I'm doing vibe coding and building a ton of software but I don't know about the vocabulary and the backend of a lot of the elements. I know that it has abilities to do tons and tons of features, including to spin up and run full AI agents on their platform, but I want to understand it at a very deep level. Research every possible article, all of the FAQ pages, all the deep pages, anything available on the domain or subdomains for Cloudflare.com and beyond, and inform me as if I know nothing. I want to know everything and become an absolute expert in all of its capabilities from a technical standpoint but then also put it in very plain English so I can understand. Demonstrate all the companies you can identify that use it. Research the articles that may be companies or spokesmen from companies that work with CloudFlare have referenced what they use it for so I can understand on the backend exactly what the functionality is. I don't know what I don't know and I want to build a lot more sophisticated software to solve problems. I think I can build the vast majority of all of it with databases and capabilities that I didn't even know existed on their platform. I need to know: - the history of it - how it's evolved - what it's capable of - the companies you can identify that use it - which of the features they use Maybe describe the examples of the feature capabilities in terms of user experience, not in technical terms of how I use a product that you run on Cloudflare. By actual customers they have, list all of its capabilities and then provide multiple examples of what their different levels of work are, storage and databases, browser run and AI agents, et cetera. I can understand the entire ecosystem of what all I can build and host. Also list all their major competitors, like Supabase, Netlify, Vercel. What all is considered competitors? What can and can't I do on Cloudflare? What do some of the competitors do? I want to understand that whole ecosystem for purposes of how I'm going to apply it for our own internal technology and marketing and sales. I'm really interested in it for marketing and sales strategy and how to use their techniques and technology. Also how we can build our own internal databases and And build software products for our clients using AI agents and architecture and pricing. What is all their pricing? How does the pricing work for it for all their different capabilities and maybe examples of what some companies are paying monthly for CloudFlare? As it seems like all their stuff is super super cheap. Even their free plan gives you a ton of ability. Why is it so cheap and all the different AI agent integrations that they already have natively on there, etc.? I just want to learn a massive amount about this company as I think it can be the foundation for a ton that we go to build moving forward at Outsource Access, supporting our clients internally. The tools we can build that we're doing right now to replace tons of Google spreadsheets and to provide dedicated login dashboard environments for our clients and for our employees and to manage workflows internally. What we can offer for our clients and how we can integrate our work with them and operate on their platform. Right now we're on Google Workspace and it has some ability to do agents, etc. I think we just want to use Google for email and Google Chat for communications and Google Drive for documents. Frankly other non-Google documents like PDFs, etc., we have to save in a secure environment and control our staff's engagement. A lot of our entire AI ecosystem I think we can build entirely on Cloudflare. Other companies I want to start entirely new software web app software applications for different things like: - creating a database tool that can help coaches make lineups for their teams - a family hub web app environment that families can store all their information but has integrated AI tools built in - personal CRM systems for individuals like I've built for myself called Brad's Command Center that's like a custom CRM built for exactly how my ADHD brain works. It integrates with all of my email and my calendar systems and pulls it in and creates dashboards and tasks lists and things exactly what I want, exactly like what I get inside of a hub spot or some of these other tools but hyper customized for what I want My biggest concern on these is the security layer. We build something either internally or external products. What is the security with Cloudflare and anything you have to add in for things like HIPAA compliance for healthcare and data security and SOC 2 type compliance? Can we get that with Cloudflare hosting? Really just move everything there instead of Google Workspace because it seems so difficult to get Google agents built and to do what I want them to do. --- # Turn 2 · Assistant **Category:** Original research **Title:** Cloudflare deep strategic and technical briefing # Cloudflare: a deep strategic and technical briefing **Current through July 22, 2026** Cloudflare changes unusually quickly. By my count, its current developer-documentation directory contains **128 named entries**, although that number includes actual products, product suites, administrative areas, APIs, and documentation collections—not 128 separately priced products. Cloudflare itself describes the commercial platform as offering more than 60 cloud services. ([developers.cloudflare.com](https://developers.cloudflare.com/directory/)) I have treated “complete” here as: - Every currently named Cloudflare product or platform surface. - The major functions inside those products. - The material corporate, funding, leadership, ownership, growth, stock, acquisition, controversy, and reliability history. - Documented customer uses. - The most important current pricing dimensions. - The practical implications for Outsource Access and the products you are considering. It would not be sensible to reproduce every individual API endpoint, configuration switch, or blog post: Cloudflare has thousands of them, and they change weekly. This briefing instead gives you the mental model needed to understand what those individual settings do. --- # 1. My conclusion for Outsource Access **Cloudflare could become the technical foundation for a very large portion of what you want to build.** It can host interfaces, APIs, workflows, databases, files, real-time systems, AI inference, AI retrieval, browser automation, agent memory, code sandboxes, client-specific environments, and much of the security perimeter. But I would **not** “move everything from Google Workspace to Cloudflare.” The best architecture is: - **Google Workspace:** email, calendar, Google Chat, collaborative documents, organizational identity. - **Cloudflare:** application hosting, client portals, internal dashboards, APIs, workflows, secure file delivery, Zero Trust access, application security, AI orchestration, network controls, and browser/code execution. - **D1 or an external PostgreSQL service:** application databases, depending on complexity. - **R2:** private PDFs, attachments, generated reports, exports, recordings, and other application files. - **Cloudflare Access with Google Workspace:** employee login and internal application access. - **An external customer-identity product when needed:** Auth0, Clerk, Supabase Auth, WorkOS, or another consumer identity provider for customers and family-app users. - **AI Gateway plus Workers AI or external AI providers:** central control over model calls, logs, cost, retries, and provider switching. Google Workspace already integrates with Cloudflare Access as an identity provider. Employees can authenticate to a Cloudflare-protected application with their Google credentials, and Access can use Google Workspace group membership when determining who is permitted to enter. Google Workspace SCIM provisioning is not presently supported by that integration, so some lifecycle management still requires additional design. ([developers.cloudflare.com](https://developers.cloudflare.com/cloudflare-one/identity/idp-integration/google-workspace/)) For healthcare, the conclusion is stricter: > **Cloudflare can be part of a HIPAA-compliant system, but simply running an application on Cloudflare does not make it HIPAA compliant.** Cloudflare offers a BAA to eligible Enterprise customers. You must confirm that every Cloudflare service that receives, stores, transmits, or logs protected health information is covered by that BAA and by Cloudflare’s current compliance scope. Your organization must still implement authorization, auditing, retention, incident response, workforce policies, backups, risk analysis, and vendor management. ([cloudflare.com](https://www.cloudflare.com/learning/privacy/what-is-hipaa-compliance/)) --- # 2. What Cloudflare actually is Cloudflare started as a reverse proxy, CDN, DNS, and security service. It has become what it calls a **connectivity cloud**: one global software-defined network that can sit between users, employees, applications, clouds, databases, AI models, and the public Internet. A normal Cloudflare application request may travel through this sequence: **User → Cloudflare DNS → TLS encryption → DDoS protection → web application firewall → bot/fraud checks → identity check → cache or routing optimization → Worker or container → database/file store/AI model → logging → response** Cloudflare says its network now operates in more than **335 cities across more than 125 countries**, reaches approximately 95% of the Internet-connected population within 50 milliseconds, blocks roughly 234 billion cyber threats daily, and has more than 500 Tbps of external network capacity. It also says 42% of the Fortune 500 are customers. ([cloudflare.com](https://www.cloudflare.com/press/press-kit/)) Its architectural advantage is that many services run across the same network and often on the same underlying fleet. Cloudflare describes the design as running every service on every server, rather than building isolated regional product silos. That makes it possible to add security, computation, caching, networking, and AI close to users without provisioning a conventional server in every region. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000016/cloud-20251231.htm)) ## The plain-English analogy Think of Cloudflare as a combination of: - The security guards outside your building. - The roads and traffic-control system that direct visitors. - A worldwide network of small offices near every customer. - A set of computers in those offices that can run your application. - File cabinets and databases attached to those computers. - An AI switchboard that can call different AI models. - A secure remote browser that can use websites for your software. - A gatekeeping system that decides which employee, customer, application, or AI agent can access each resource. That is why Cloudflare competes with portions of so many different companies. --- # 3. Is “20% of the Internet runs on Cloudflare” accurate? It is directionally meaningful but imprecise. Cloudflare’s current language includes both “powering 20% of the Internet” and “20% of all websites are protected by Cloudflare.” The latter is the more defensible interpretation. It does **not** mean Cloudflare owns or hosts 20% of all servers, Internet traffic, data, or applications. ([cloudflare.com](https://www.cloudflare.com/)) W3Techs, which examines detectable technologies used by public websites, reported on July 22, 2026 that Cloudflare was used by approximately **24.1% of all websites** it measured and by approximately **84% of the websites whose reverse-proxy service it could identify**. W3Techs’ methodology measures public websites and technologies detectable from the outside; it is not a census of all Internet bytes or infrastructure. ([w3techs.com](https://w3techs.com/technologies/overview/proxy)) A careful statement would therefore be: > Cloudflare sits in front of or supplies detectable services to roughly one-fifth to one-quarter of public websites, depending on the dataset and measurement date. That is still an extraordinary footprint. It gives Cloudflare enormous visibility into attacks, bots, routing conditions, crawler behavior, browser behavior, and global Internet performance. --- # 4. How Cloudflare began ## Project Honey Pot: 2004 Matthew Prince and Lee Holloway originally created **Project Honey Pot** to understand where email spam came from. Websites participating in the project could help identify the IP addresses and techniques used by spammers to harvest email addresses. Over time, participating website owners repeatedly asked for a service that would stop abusive traffic rather than merely report it. ([cloudflare.com](https://www.cloudflare.com/our-story/)) ## The Cloudflare concept: 2009 While attending Harvard Business School, Prince met Michelle Zatlyn. Zatlyn saw that Project Honey Pot’s data and community could become a service protecting and accelerating websites. Prince, Zatlyn, and Holloway became Cloudflare’s three co-founders, with Holloway supplying much of the original technical architecture. The early working name was “Project Web Wall”; the company eventually adopted Cloudflare. ([cloudflare.com](https://www.cloudflare.com/our-story/)) Cloudflare participated in the Harvard Business School business-plan competition in April 2009 and raised an initial round later that year. Its earliest service was tested with members of the Project Honey Pot community. ([cloudflare.com](https://www.cloudflare.com/our-story/)) ## Public launch: September 27, 2010 Cloudflare launched publicly at TechCrunch Disrupt on September 27, 2010. Its original proposition was unusually simple: 1. Change your domain’s DNS settings. 2. Cloudflare would sit between your visitors and your existing server. 3. Malicious traffic would be filtered. 4. legitimate pages would load faster. 5. The basic service would be free. The original team had only a handful of people, and traffic reportedly doubled weekly during the early launch period. ([blog.cloudflare.com](https://blog.cloudflare.com/page/176/)) --- # 5. The strategic evolution of Cloudflare | Period | What changed | Strategic significance | |---|---|---| | **2004–2009** | Project Honey Pot collected abuse and reputation data. | Created the threat-data and community foundation. | | **2009–2010** | Cloudflare was formed, funded, privately tested, and publicly launched. | Converted threat intelligence into a preventive network service. | | **2011–2013** | IPv6 support, rapid domain growth, additional venture funding, and international network expansion. | Established Cloudflare as a scalable CDN/security network rather than a small filtering service. | | **2014** | Universal SSL made HTTPS available to Cloudflare sites without a separate certificate purchase. | Helped make encryption a default feature rather than a premium add-on. | | **2015–2016** | Major strategic funding and international expansion; spelling changed from “CloudFlare” to “Cloudflare.” | Prepared the company to compete with much larger infrastructure providers. | | **2017** | Unmetered DDoS protection, Stream, Geo Key Manager, and Workers were introduced. | This was the critical transition from “website protection” to “programmable cloud platform.” | | **2018** | 1.1.1.1, Registrar, Bandwidth Alliance, Workers KV, and additional developer services. | Added consumer privacy, domains, storage, and a stronger developer ecosystem. | | **2019** | WARP, additional bot and performance services, Workers Sites, and the IPO. | Became a public company and broadened from enterprise infrastructure to consumers and developers. | | **2020** | Browser Isolation through S2 Systems, Cloudflare Teams/Zero Trust development, and Cloudflare Radar. | Entered enterprise workforce security and Internet intelligence. | | **2021** | Magic WAN, network firewall capabilities, Zaraz, and deeper Zero Trust. | Began replacing traditional corporate networks, firewalls, VPNs, and tag managers. | | **2022** | R2 object storage, Queues, Area 1 Email Security, Vectrix/CASB, and related services. | Expanded into storage, asynchronous applications, email protection, and SaaS security. | | **2023** | Workers AI, AI Gateway, and Vectorize were introduced. | Established the first version of Cloudflare’s AI development stack. | | **2024** | Workers AI and Vectorize reached general availability; Baselime and Kivera were acquired. | Improved AI production readiness, observability, and cloud-security governance. | | **2025** | Agents SDK, AI Search/AutoRAG, Containers, Browser Rendering expansion, Data Platform, Outerbase, and Replicate. | Moved from individual AI primitives toward an integrated AI-agent cloud. | | **2026** | Agent Memory, Artifacts, Dynamic Workers, Sandboxes, Mesh, Workers VPC, unified AI inference, Email Service, Flagship, Agent Lee, Precursor, AI-content controls, and major developer-tooling acquisitions. | Cloudflare is now positioning itself as an operating environment for autonomous agents, applications, people, and content economics. | The early milestones are documented in Cloudflare’s own history and product timeline. ([cloudflare.com](https://www.cloudflare.com/our-story/)) The later developer and AI milestones are documented in Cloudflare’s Workers AI, Agents, Containers, Workflows, Browser Run, Data Platform, and 2026 Agents Week announcements. ([blog.cloudflare.com](https://blog.cloudflare.com/workers-ai-ga-huggingface-loras-python-support/)) --- # 6. Funding and investment history Cloudflare was venture-funded before becoming public. It was not built as a conventional private-equity-controlled rollup. | Approximate date | Amount | Notable participants and purpose | |---|---:|---| | **November 2009** | Just over **$2 million** | Venrock and Pelion Venture Partners; funded initial product and network development. | | **July 2011** | **$20 million** | Led by New Enterprise Associates with existing investors. | | **December 2012** | **$50 million** | Led by Union Square Ventures, with Greenspring and existing investors. | | **2014–2015** | **$110 million** | Included Fidelity, Baidu, Google/CapitalG, Microsoft, and Qualcomm interests. This brought strategic cloud, mobile, and international relationships. | | **2019 pre-IPO** | **$150 million** | Led by Franklin Templeton; supported expansion, product development, and international growth. | | **September 2019 IPO** | **$15 per share** | Cloudflare sold 35 million shares, with an underwriters’ option for additional shares. | Cloudflare reported more than $330 million raised across five private funding rounds before the IPO. ([blog.cloudflare.com](https://blog.cloudflare.com/growing-cloudflare/)) ## Was Cloudflare funded or controlled by private equity? **No private-equity firm controls Cloudflare.** Its pre-IPO backers were primarily: - Venture-capital firms. - Mutual-fund and asset-management organizations. - Corporate strategic investors. - Growth investors. Following the IPO, institutional investors own economically meaningful portions of the public shares, but the founders retain disproportionate voting power through a dual-class share structure. A board member’s employment by a private-equity firm does not make that private-equity firm Cloudflare’s owner. For example, Stacey Cunningham’s role at Advent is distinct from Cloudflare’s ownership structure. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000119312526263809/d160227ddef14a.htm)) --- # 7. Acquisitions: how Cloudflare buys capabilities Cloudflare’s acquisition pattern is generally **capability-led** rather than revenue-rollup-led. It often buys a small company or team whose technology can become a native service running across the Cloudflare network. | Date | Acquisition or team | What it added | |---|---|---| | **2014** | StopTheHacker | Website malware detection and remediation. | | **2014** | CryptoSeal | VPN, encryption, and trusted-computing expertise. | | **2016** | Eager | Easier installation and distribution of website applications. | | **2017** | Neumob | Mobile application performance technology. | | **2020** | S2 Systems | Remote Browser Isolation, foundational to Cloudflare One. | | **2021** | Zaraz | Edge-based management of analytics, advertising, and third-party web scripts. | | **2022** | Vectrix | SaaS security posture and CASB capabilities. | | **2022** | Area 1 Security | Email phishing, business-email-compromise, and malicious-message detection. | | **2024** | Baselime | Serverless observability and developer debugging. | | **2024** | Kivera | Cloud-security, policy, data-protection, and compliance technology. | | **2025** | Outerbase | Database developer experience and AI-assisted database tooling. | | **2025** | Replicate | Simplified deployment and operation of AI models; acquisition completed for approximately $57.4 million in cash. | | **2026** | Human Native | AI-content licensing and a marketplace approach to creator/model relationships. | | **2026** | Astro team | Content-oriented web framework talent and frontend-development capability. | | **2026** | VoidZero team | Vite, Vitest, Rolldown, Oxc, and related JavaScript tooling expertise. | | **2026** | Ensemble AI talent | Additional AI research and product capability. | ([blog.cloudflare.com](https://blog.cloudflare.com/cloudflare-acquires-stopthehacker/)) The pattern matters: Cloudflare frequently takes a stand-alone category—browser isolation, email security, serverless observability, AI model deployment—and integrates it into the same network and account model. --- # 8. Leadership and leadership evolution ## Current executive leadership | Person | Current role | Significance | |---|---|---| | **Matthew Prince** | Co-founder, CEO, and Co-Chair | Strategy, external voice, network vision, corporate mission, and capital-market leadership. | | **Michelle Zatlyn** | Co-founder, President, and Co-Chair | Product, operations, customers, strategy, and organizational leadership. | | **Thomas Seifert** | Chief Financial Officer | Finance, capital allocation, investor relations, and public-company operations. | | **Dane Knecht** | Chief Technology Officer | Longtime Cloudflare leader overseeing technology and product direction, including developer and AI platforms. | | **Grant Bourzikas** | Chief Security Officer | Enterprise and internal security strategy. | | **Mark Anderson** | President of Revenue | Sales, enterprise expansion, and commercial operations. | | **Alissa Starzak** | Chief Legal Officer and Secretary | Legal, regulatory, policy, governance, and corporate-secretary functions as of April 2026. | ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000026/cloud-20251231.htm)) ## Important leadership changes - Michelle Zatlyn’s role evolved from co-founder and operating leader to President and Co-Chair. - Matthew Prince and Michelle Zatlyn became board Co-Chairs in February 2025. - John Graham-Cumming, a central early engineering leader and CTO from 2016, moved to Cloudflare’s board in 2025. - Dane Knecht became CTO after a long tenure leading major product areas. - Scott Sandell became lead independent director. - Alissa Starzak succeeded Douglas Kramer as Chief Legal Officer in April 2026. - The board added John Graham-Cumming, Stacey Cunningham, and Harvard professor Karim Lakhani in 2025, increasing technical, operating, financial-market, and AI expertise. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000119312526263809/d160227ddef14a.htm)) Cloudflare’s SEC filings explicitly acknowledge dependence on Prince and Zatlyn and the importance of integrating newer senior leaders. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000016/cloud-20251231.htm)) --- # 9. Ownership and founder control Cloudflare has Class A and Class B shares: - Class A shares generally receive one vote per share. - Class B shares generally receive ten votes per share. As of April 30, 2026, SEC disclosures showed approximately: - Matthew Prince: **39.0% of total voting power**. - Michelle Zatlyn: **13.4% of total voting power**. - Executive officers and directors as a group: approximately **52.4% of total voting power**. Major disclosed Class A holders included Capital World Investors, Baillie Gifford, Morgan Stanley, BlackRock, and Vanguard entities. These institutions generally manage shares for funds and clients; they are not the equivalent of a controlling private-equity owner. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000119312526263809/d160227ddef14a.htm)) Cloudflare also proposed—and shareholders approved in 2026—governance changes involving non-voting Class C shares. The stated purpose included preserving long-term founder-led direction while allowing the company to issue equity for employees and acquisitions without the same voting dilution. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000119312526263809/d160227ddef14a.htm)) **Practical interpretation:** public shareholders own the economics, but Prince and Zatlyn retain unusual control over strategic direction. --- # 10. Revenue, customers, employees, and growth ## Revenue history | Year | Revenue | |---:|---:| | 2016 | $84.8 million | | 2017 | $134.9 million | | 2018 | $192.7 million | | 2019 | $287.0 million | | 2020 | $431.1 million | | 2021 | $656.4 million | | 2022 | $975.2 million | | 2023 | $1.297 billion | | 2024 | $1.670 billion | | 2025 | $2.168 billion | That is a compound annual growth rate of approximately **43% from 2016 through 2025**, calculated from Cloudflare’s reported revenue. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733321000009/cloud-20201231.htm)) For 2025: - Revenue increased approximately 30%. - Gross margin was approximately 75%. - Sales and marketing consumed approximately 43% of revenue. - Research and development consumed approximately 24%. - Cloudflare reported an operating loss and a net loss of approximately $102 million. - R&D spending was approximately $512.5 million. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000016/cloud-20251231.htm)) For the first quarter of 2026: - Revenue was approximately **$639.8 million**, up 34% year over year. - GAAP operating loss was approximately $62 million. - Non-GAAP operating income was approximately $73.1 million. - Cash, equivalents, and marketable securities were approximately $4.16 billion. - Free cash flow was approximately $84.1 million. ([cloudflare.com](https://www.cloudflare.com/press/press-releases/2026/cloudflare-announces-first-quarter-2026-financial-results/)) Cloudflare’s 2026 full-year revenue guidance following that quarter was approximately **$2.805 billion to $2.813 billion**. ([cloudflare.com](https://www.cloudflare.com/press/press-releases/2026/cloudflare-announces-first-quarter-2026-financial-results/)) ## Customer growth | Metric | 2023 | 2024 | 2025 | |---|---:|---:|---:| | Paying customers | 189,791 | 237,714 | 332,466 | | Customers generating over $100,000 in annualized revenue | 2,756 | 3,497 | 4,298 | | Dollar-based net retention | 115% | 111% | 120% | ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000016/cloud-20251231.htm)) The number of paying customers grew almost 40% in 2025, while the number of $100,000-plus customers grew approximately 23%. Net retention of 120% means the previous customer cohort, taken together, was spending about 20% more after accounting for expansion and contraction. ## Employee history | Date | Reported employees | |---|---:| | 2019 | Approximately 1,270 | | 2020 | Approximately 1,788 | | 2021 | Approximately 2,439 | | 2022 | Approximately 3,217 | | 2023 | Approximately 3,682 | | 2024 | Approximately 4,263 | | December 31, 2025 | **5,156** | ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733320000010/cloud-20191231.htm)) In May 2026, Cloudflare announced a planned reduction of approximately **1,100 positions**, around 20% of the then-reported workforce, with restructuring expected to be substantially completed by the third quarter. Cloudflare linked the restructuring to organizational redesign and increased internal AI use. The exact July 2026 employee count is not publicly established; subtracting 1,100 from the prior audited count would be only a mechanical estimate and would ignore hiring and normal attrition. ([cloudflare.com](https://www.cloudflare.com/press/press-releases/2026/cloudflare-announces-first-quarter-2026-financial-results/)) --- # 11. Stock performance and valuation pattern Cloudflare trades on the New York Stock Exchange under **NET**. - IPO price in September 2019: **$15 per share**. - First-day closing price: approximately **$17.90**. - Latest quote returned for July 22, 2026: approximately **$268.98**. - Market capitalization at that quote: approximately **$94.9 billion**. - The stock was still reporting negative trailing GAAP earnings per share, so a conventional positive P/E ratio was not meaningful. ([cloudflare.com](https://www.cloudflare.com/press/press-releases/2019/cloudflare-announces-pricing-of-initial-public-offering/)) The long-term stock pattern has been: 1. Strong post-IPO appreciation. 2. An extreme cloud-software valuation expansion during 2020–2021. 3. A major contraction during the 2022 technology selloff. 4. Recovery as revenue growth stayed high and Cloudflare expanded into Zero Trust, developer infrastructure, and AI. 5. New highs in July 2026 amid optimism about the agentic-cloud strategy. ([cloudflare.net](https://cloudflare.net/stock-information/Stock-Quote--Chart/default.aspx)) At approximately $94.9 billion of market capitalization against $2.17 billion of 2025 revenue, the simple trailing market-cap-to-revenue ratio was around **44 times**. This is not a complete valuation analysis, but it shows that the market is pricing in prolonged high growth, expanding margins, and significant AI/cloud-platform success. Major stock risks include: - High valuation expectations. - Continued GAAP losses. - Gross-margin pressure as GPU and compute workloads grow. - Competition from hyperscalers and specialized platforms. - Dependence on continued enterprise expansion. - Reliability or security failures. - The possibility that its AI products take longer to monetize than investors expect. --- # 12. Cloudflare’s core business pattern Cloudflare’s most important strategic pattern is **land and expand**: 1. A customer starts with DNS, CDN, free SSL, or basic DDoS protection. 2. Cloudflare adds WAF, bot management, rate limiting, load balancing, and logs. 3. The customer adds employee Access, Gateway, CASB, DLP, and email security. 4. Developers add Workers, Pages, R2, D1, Queues, or Durable Objects. 5. The organization adds AI Gateway, Workers AI, AI Search, agents, browser automation, and agent security. 6. The customer signs a larger, multi-product enterprise agreement. Cloudflare says most customers use free services and most free users never convert. But the free plan supplies brand awareness, developer adoption, Internet-scale traffic diversity, operational learning, and a pipeline from which some users become meaningful paid customers. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000016/cloud-20251231.htm)) Enterprise customers increasingly purchase a committed “pool of funds” that may be consumed across multiple products rather than paying a clean retail rate for each individual service. This is why named companies’ exact monthly Cloudflare bills are rarely public. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000016/cloud-20251231.htm)) --- # 13. Complete current product and capability map The following preserves all 128 names in Cloudflare’s current documentation directory. Some are product families or administrative surfaces rather than separately billable services. The descriptions are plain-English translations of their purpose. ([developers.cloudflare.com](https://developers.cloudflare.com/directory/)) ## A. Global network, DNS, caching, and performance | Surface | What it does in plain English | |---|---| | **1.1.1.1** | A public DNS resolver intended to make Internet lookups fast and private; also associated with the consumer WARP application. | | **Argo Smart Routing** | Measures Internet conditions and routes traffic around congestion or failures. | | **Automatic Platform Optimization** | Caches and accelerates WordPress pages, including dynamic HTML. | | **Bring Your Own IP** | Lets Cloudflare advertise and protect IP-address ranges that your organization owns. | | **Cache** | Stores copies of content close to users so the origin server is contacted less often. | | **Cache Reserve** | Keeps a persistent R2-backed copy of cacheable content, reducing origin requests and egress. | | **China Network** | Extends application delivery and security into mainland China through partner infrastructure. | | **DNS** | Authoritative domain-name service: the Internet’s address book for your domains. | | **Google Tag Gateway** | Serves Google tags through a first-party path to improve performance and resilience. | | **Health Checks** | Continuously tests whether an origin server or endpoint is working. | | **Load Balancing** | Sends users to healthy servers or regions according to geography, load, latency, or policy. | | **Network** | Umbrella documentation for Cloudflare’s network platform and connectivity services. | | **Network Error Logging** | Receives browser reports about connection and delivery failures. | | **Registrar** | Registers and renews domains with pricing intended to reflect wholesale registry cost. | | **Smart Shield** | Bundles tiered cache, connection reuse, origin protection, and optionally Argo and Cache Reserve. | | **Speed** | Measures and optimizes page-loading performance. | | **Waiting Room** | Places excess visitors into a controlled virtual queue rather than allowing a site to crash. | Smart Shield now combines tiered caching, connection reuse, health monitoring, Cache Reserve, and optional Argo capabilities. ([developers.cloudflare.com](https://developers.cloudflare.com/smart-shield/)) ## B. Application, API, bot, fraud, and content security | Surface | What it does | |---|---| | **AI Crawl Control** | Identifies and controls AI crawlers; allows policies around search, model input, training, and monetization. | | **API Shield** | Protects APIs with schema validation, client certificates, authentication checks, discovery, and abuse controls. | | **Bots** | Distinguishes humans, beneficial bots, automated abuse, scrapers, and sophisticated bot traffic. | | **Challenges** | Performs browser or behavior checks when traffic looks suspicious. | | **Client-side security** | Monitors JavaScript and browser-side resources for tampering, supply-chain attacks, or data leakage. | | **DDoS protection** | Automatically absorbs network- and application-layer traffic floods. | | **DMARC Management** | Helps monitor email authentication and domain spoofing. | | **Fraud Detection** | Detects signals associated with abusive accounts, transactions, or automated fraud. | | **Geo Key Manager** | Controls the geographic regions in which TLS private keys may be stored or used. | | **Keyless SSL** | Lets Cloudflare terminate encrypted connections while the private key remains in your own HSM or key server. | | **Leaked Credentials Detection** | Detects login attempts using credentials known to have appeared in data breaches. | | **Rate Limiting** | Restricts how frequently a user, IP, token, or endpoint can perform an action. | | **Security Center** | Central asset, risk, exposure, and security-findings view. | | **SSL/TLS** | Manages HTTPS, certificates, encryption modes, cipher policies, and origin encryption. | | **Turnstile** | A CAPTCHA alternative that evaluates whether a request is legitimate without routinely asking users to solve puzzles. | | **WAF** | Applies managed and custom rules to stop attacks such as injection, exploit attempts, and application abuse. | ## C. Zero Trust, SASE, and employee security | Surface | What it does | |---|---| | **Access** | Replaces traditional VPN access to internal applications with identity- and device-aware authorization. | | **Browser Isolation** | Runs an employee’s browsing session remotely so malicious web code does not execute directly on the device. | | **CASB** | Finds risky SaaS configurations, oversharing, exposed data, and unsanctioned cloud applications. | | **Cloudflare One** | Umbrella SASE platform combining employee access, network security, email, data, and SaaS controls. | | **Cloudflare One Appliance** | Physical or virtual on-ramp connecting offices and networks to Cloudflare One. | | **Cloudflare One Client** | Unified endpoint software for routing, posture checking, DNS filtering, and secure access. | | **Cloudflare Tunnel** | Creates an outbound-only encrypted connection from your environment to Cloudflare, avoiding an exposed inbound server port. | | **Cloudflare Tunnel for SASE** | Uses tunnels specifically to connect private applications and networks into Cloudflare One. | | **Cloudflare WAN** | Software-defined wide-area networking for offices, data centers, and clouds. | | **DLP** | Detects and restricts movement of sensitive information such as health, identity, payment, or proprietary data. | | **Digital Experience Monitoring** | Measures whether employee applications, devices, networks, and SaaS services are performing properly. | | **Email Security** | Detects phishing, malicious links, malicious attachments, impersonation, and business-email compromise. | | **Gateway** | Filters employee DNS, HTTP, network, and SaaS traffic according to security policy. | | **Internal DNS** | Provides consistent private-name resolution across offices, remote employees, and clouds. | | **Public-to-private connectivity** | Lets a user address a familiar hostname while Cloudflare privately routes the request to a non-public service. | | **WARP** | Cloudflare’s endpoint traffic-routing client, used for consumer privacy and enterprise Zero Trust connectivity. | **Browser Isolation and Browser Run are different:** Browser Isolation protects a human who is browsing; Browser Run gives your software or AI agent a programmable browser. ## D. Network security and multicloud connectivity | Surface | What it does | |---|---| | **Mesh** | Identity-aware connectivity among people, applications, services, AI agents, and multiple clouds. | | **Network Firewall** | Applies network-layer policies to IP traffic crossing Cloudflare. | | **DNS Firewall** | Provides protected recursive DNS services for corporate networks and service providers. | | **Magic Transit** | Advertises and protects a customer’s IP network from DDoS attacks while routing legitimate traffic onward. | | **Multi-Cloud Networking** | Connects services distributed across AWS, Azure, Google Cloud, private clouds, and data centers. | | **Network Flow** | Analyzes NetFlow, IPFIX, sFlow, and cloud-flow data for traffic visibility, capacity planning, and attack detection. | | **Network Interconnect** | Provides private physical or virtual connections between customer networks and Cloudflare. | | **Spectrum** | Proxies and protects non-HTTP TCP or UDP applications, such as games, SSH, messaging, and custom protocols. | | **Workers VPC** | Gives Workers private connectivity to databases and services that should not be exposed to the public Internet. | | **Data Localization Suite** | Controls where traffic is inspected, encryption keys are used, and selected logs or metadata are processed. | Network Flow was formerly called Magic Network Monitoring and can ingest traditional network-flow protocols and supported cloud flow logs. ([developers.cloudflare.com](https://developers.cloudflare.com/network-flow/)) Mesh, introduced in 2026, extends Cloudflare’s identity-aware networking model to humans, services, clouds, and AI agents. ([cloudflare.com](https://www.cloudflare.com/press/press-releases/2026/cloudflare-launches-mesh-to-secure-the-ai-agent-lifecycle/)) ## E. Application hosting, compute, and execution | Surface | What it does | |---|---| | **Cloudflare for Platforms** | Umbrella toolkit for companies that want to build Cloudflare infrastructure into their own product. | | **Cloudflare for SaaS** | Gives each SaaS customer a custom hostname, certificate, routing, and security configuration. | | **Containers** | Runs conventional containerized software, including applications requiring binaries, packages, or longer-lived processes. | | **Durable Objects** | Creates millions of individually addressable, stateful “rooms” or actors for coordination, real-time state, and local SQL storage. | | **Dynamic Workers** | Creates isolated Workers from code supplied at runtime, including untrusted or AI-generated code. | | **Email Service** | Sends email through a Workers binding, API, or SMTP and processes inbound email programmatically. | | **Flagship** | Provides feature flags, targeted releases, percentage rollouts, and instant feature control without redeploying code. | | **Pages** | Deploys frontend sites and associated server-side functions from Git or build output. | | **Sandbox SDK** | Creates isolated Linux execution environments for agents, builds, shell commands, files, and background processes. | | **Secrets Store** | Keeps API keys, credentials, and secrets outside application source code. | | **Workers** | Runs serverless application code across Cloudflare’s global network. | | **Workers for Platforms** | Allows a platform to run isolated code on behalf of many customers or tenants. | | **Workflows** | Runs durable, multi-step processes that survive delays, failures, restarts, and external-service interruptions. | | **Artifacts** | Stores versioned file trees through a Git-compatible interface, particularly for coding agents and automated builds. | Dynamic Workers can be created at runtime, can be restricted to selected bindings and network access, and are intended as a lightweight sandbox for code that should not be fully trusted. ([developers.cloudflare.com](https://developers.cloudflare.com/dynamic-workers/)) Flagship is Cloudflare’s native OpenFeature-compatible feature-flag service. ([developers.cloudflare.com](https://developers.cloudflare.com/flagship/)) Email Service can send through Workers, a REST API, or authenticated SMTP and can process inbound mail with Worker logic. ([developers.cloudflare.com](https://developers.cloudflare.com/email-service/api/route-emails/email-handler/)) Artifacts is currently in closed beta and creates isolated, durable, Git-compatible repositories for agents, users, branches, or tasks. ([developers.cloudflare.com](https://developers.cloudflare.com/artifacts/)) ## F. Databases, object storage, data platforms, and asynchronous processing | Surface | What it does | |---|---| | **D1** | Managed, serverless SQL databases based on SQLite. | | **Hyperdrive** | Pools and accelerates connections from Workers to external PostgreSQL or MySQL databases. | | **KV** | Globally distributed key-value storage for settings, configuration, cached data, sessions, and read-heavy workloads. | | **Pipelines** | Receives and transforms streams of events or data before writing them into R2. | | **Queues** | Stores messages until a background consumer is ready to process them reliably. | | **R2** | S3-compatible object storage for documents, images, backups, datasets, recordings, and large files, without normal Internet-egress charges. | | **R2 Data Catalog** | Adds a structured data-lake catalog to datasets stored in R2. | | **R2 SQL** | Queries data-lake tables stored in R2 using SQL. | | **Vectorize** | Stores embeddings for semantic search, recommendation, similarity, and retrieval-augmented AI. | | **Analytics Engine** | Records high-cardinality application events and allows aggregate analytics without operating a traditional analytics database. | ## G. AI, agents, retrieval, and browser automation | Surface | What it does | |---|---| | **Agent Lee** | Cloudflare’s own account-aware AI assistant for diagnosing and configuring Cloudflare, with approvals before changes. | | **Agent Memory** | Managed long-term memory that extracts and recalls useful facts for an agent, with namespaces for tenant isolation. | | **Agents** | SDK and runtime for stateful, real-time, tool-using AI agents. | | **AI** | Umbrella documentation for Cloudflare’s AI platform. | | **AI Gateway** | Central gateway to models from Cloudflare or external providers, adding logs, retries, caching, budgets, analytics, and provider routing. | | **AI Search** | Ingests documents and creates hybrid keyword-and-semantic retrieval for grounded AI answers. | | **Browser Run** | Gives code or an agent a managed Chromium browser for navigation, extraction, screenshots, PDFs, testing, and browser tasks. | | **Workers AI** | Runs supported text, image, embedding, speech, and other AI models on Cloudflare’s GPU network. | ## H. Media, images, voice, video, and real-time communications | Surface | What it does | |---|---| | **Images** | Uploads, stores, resizes, transforms, optimizes, and delivers images. | | **Media over QUIC** | Implements emerging low-latency media delivery over QUIC. | | **Realtime** | Umbrella platform for real-time audio, video, and communication. | | **SFU** | Selective forwarding infrastructure for multiparty real-time video and audio. | | **RealtimeKit** | Higher-level SDKs and components for embedding calls, meetings, and interactive media. | | **Stream** | Uploads, encodes, stores, protects, broadcasts, and delivers live or recorded video. | | **TURN** | Relays WebRTC traffic when two participants cannot establish a direct connection. | ## I. Analytics, logs, and Internet intelligence | Surface | What it does | |---|---| | **Analytics** | Umbrella reporting for traffic, security, performance, applications, and products. | | **GraphQL Analytics API** | Programmatically queries detailed Cloudflare analytics datasets. | | **Log Explorer** | Searches and investigates retained Cloudflare logs. | | **Logs** | Captures, filters, stores, and exports traffic and product events. | | **Radar** | Publishes aggregated information about Internet traffic, outages, attacks, protocols, AI bots, and global trends. | | **Web Analytics** | Privacy-oriented website visitor and performance analytics. | ## J. Privacy, cryptographic trust, and public Internet utilities | Surface | What it does | |---|---| | **Key Transparency Auditor** | Audits transparency systems used to verify public encryption keys. | | **Privacy Gateway** | Infrastructure for privacy-preserving request relays and oblivious HTTP patterns. | | **Privacy Pass** | Issues anonymous cryptographic tokens that let a user prove legitimacy without exposing identity. | | **Privacy Proxy** | Tools for building privacy-preserving proxy and relay services. | | **Randomness Beacon** | Publishes verifiable public randomness for cryptographic or research uses. | | **Time Services** | Network time services, including secure time synchronization. | | **Web3** | Gateways and services for decentralized protocols such as IPFS and Ethereum. | ## K. Rules, administration, deployment, and developer management | Surface | What it does | |---|---| | **Account** | Users, memberships, roles, permissions, and account-level settings. | | **API documentation** | REST APIs for automating Cloudflare configuration and resources. | | **Billing** | Plans, subscriptions, usage, invoices, and payment management. | | **Fundamentals** | Core concepts such as zones, proxied DNS, accounts, certificates, and Cloudflare’s request path. | | **Learning Paths** | Guided documentation for common implementation goals. | | **Notifications** | Configures alerts for attacks, failures, usage, billing, and system events. | | **OAuth** | Authentication and authorization flows for connected applications and Cloudflare APIs. | | **Pulumi** | Infrastructure-as-code integration using general-purpose programming languages. | | **Reference Architecture** | Recommended designs for common application, network, security, and AI systems. | | **Resource Tagging** | Organizes Cloudflare resources by project, environment, department, client, or other metadata. | | **Rules** | Expresses traffic logic such as redirects, rewrites, cache policies, configuration changes, and routing. | | **Ruleset Engine** | Common execution framework used by WAF, redirects, transforms, and other rule-driven products. | | **Support** | Documentation and processes for account and technical assistance. | | **Tenant** | Capabilities for partners and platforms managing many subordinate customer accounts. | | **Terraform** | Manages Cloudflare configuration as version-controlled infrastructure code. | | **Version Management** | Controls application versions, staged deployments, and rollbacks. | ## L. Third-party website code and marketing tooling | Surface | What it does | |---|---| | **Zaraz** | Moves analytics, advertising, and marketing-tool execution away from the visitor’s browser and toward Cloudflare’s edge, reducing client-side scripts and improving control. | --- # 14. The most important developer primitives in plain English ## Workers: the application brain A Worker is code that runs when someone or something makes a request. It can: - Render a webpage. - Receive a form. - Authenticate a user. - Query a database. - call an AI model. - Rewrite or redirect a request. - Resize an image. - Process an API webhook. - generate a PDF or download. - enforce client-specific rules. - invoke a workflow. - send a message into a queue. - stream a response. - create a Durable Object. - call a container or browser. Workers support JavaScript, TypeScript, WebAssembly, and selected language/runtime compatibility. Node.js support has expanded but is not identical to operating an unrestricted Node server. On the paid plan, a normal invocation can use up to five minutes of CPU, with a default limit of 30 seconds; wall-clock time waiting for network responses is treated differently. ([developers.cloudflare.com](https://developers.cloudflare.com/workers/platform/limits/)) **User experience example:** an employee opens `operations.outsourceaccess.com`, Cloudflare checks their Google identity, the Worker retrieves that employee’s clients and tasks, and the page appears. ## Pages: the website or frontend Pages builds and deploys websites from source code or build files. Pages Functions are Workers attached to the site. Pages is useful for: - Marketing sites. - dashboards. - client portals. - React, Astro, Vue, or similar applications. - documentation. - prototypes and small products. Cloudflare is increasingly converging Pages and Workers so that Workers becomes the broader application platform. ## D1: the structured application database D1 is serverless SQLite. Use it for: - Clients. - employees. - tasks. - workflow statuses. - forms. - rosters. - events. - approvals. - simple CRM records. - permissions and app metadata. - moderate reporting. D1 is not PostgreSQL. It is excellent when the data model and query patterns are understandable, but it is not automatically the right choice for a giant financial system, complex data warehouse, or application dependent on PostgreSQL-specific extensions. **User experience example:** when a manager assigns an employee to a client, a row is inserted into an assignment table. Every dashboard subsequently reflects that assignment. ## R2: the private file cabinet R2 stores objects rather than database rows. Use it for: - PDFs. - contracts. - onboarding documents. - training videos. - call recordings. - screenshots. - profile photographs. - generated reports. - exports. - backups. - datasets. - AI source documents. An R2 bucket can be private. A Worker can check the user’s identity and authorization before returning a file or issuing a short-lived signed URL. R2 encrypts stored data and uses TLS in transit. ([developers.cloudflare.com](https://developers.cloudflare.com/r2/reference/data-security/)) **User experience example:** an employee sees a client SOP, but the browser never receives a permanent public URL. The app verifies the employee’s role and then provides temporary access. ## KV: the global settings cabinet KV is optimized for information that is read extremely often and changed less frequently. Good uses: - Feature settings. - tenant configuration. - cached API responses. - user preferences. - lookup tables. - temporary session-like information. - routing maps. KV is eventually consistent across locations. It should not be the authoritative system for bank balances, inventory decrements, or competing updates that require immediate global agreement. ## Durable Objects: one authoritative room for each entity A Durable Object is a globally addressable unit of state and computation. You can create one for: - Every chat room. - every live dashboard. - every family. - every game. - every collaborative document. - every client. - every AI agent. - every workflow coordinator. - every device. It can maintain WebSocket connections, run timers, and store data in local SQLite. Cloudflare routes all activity for that object to its authoritative instance, preventing conflicting updates. **User experience example:** two coaches edit a lineup simultaneously. The team’s Durable Object decides the order of changes and broadcasts the updated lineup to both screens. ## Hyperdrive: connect to a conventional database Hyperdrive accelerates and pools connections from Workers to PostgreSQL or MySQL. This is useful when: - You want Cloudflare for the application but Supabase, Neon, PlanetScale, AWS, or another provider for the database. - You need PostgreSQL features D1 does not provide. - You already have an existing database. - You want Cloudflare’s security and edge runtime without migrating everything. ## Queues: “do this later” A queue separates an immediate user request from background work. Example: 1. Employee uploads 500 documents. 2. The web request immediately confirms the upload. 3. Each document is added to a queue. 4. Background consumers extract text, classify it, create embeddings, and update the database. 5. Failures retry without forcing the employee to remain on the page. ## Workflows: “complete this reliable process” A Workflow models a sequence that may include waiting, retrying, approvals, and external systems. Example client onboarding workflow: 1. Create client record. 2. Ask manager for approval. 3. Provision folders and permissions. 4. Send welcome email. 5. wait for signed agreement. 6. import source documents. 7. create AI-search index. 8. notify assigned staff. 9. record audit completion. A Workflow survives process restarts and external-service errors. Queues are best for units of background work; Workflows are best for a stateful business process. ## Containers: conventional software without a conventional server fleet Containers support applications that need: - Native binaries. - Python or another full runtime. - long-running libraries. - operating-system packages. - more memory. - build tools. - software incompatible with Workers. They can scale down when inactive, but they introduce more startup time and cost than a small Worker. ## Sandbox SDK: an isolated workbench Sandboxes provide isolated Linux environments where an agent or application can: - Run shell commands. - install packages. - read and write files. - build a project. - run tests. - start a background process. - execute less-trusted code. This is useful for AI coding agents, data transformations, custom plugins, and client-supplied scripts. ## Dynamic Workers: extremely lightweight isolated code Dynamic Workers can execute code supplied at runtime and expose only capabilities you explicitly permit. For example, a customer-created automation could be allowed to: - Read only that customer’s database partition. - send a message to a controlled channel. - write to a restricted R2 prefix. - call one internal API. - have no arbitrary Internet access. That is a powerful foundation for safely letting clients or AI agents generate their own automations. --- # 15. The Cloudflare AI and agent stack Cloudflare does not merely provide “an AI agent product.” It provides the layers from which an agent system can be assembled. ## Layer 1: model inference — Workers AI Workers AI runs supported AI models on Cloudflare’s GPU infrastructure. Capabilities include: - Text generation. - embeddings. - image generation. - speech-to-text. - text-to-speech. - vision. - classification and other model tasks. Cloudflare has also integrated third-party model ecosystems and acquired Replicate, expanding its model availability. ([blog.cloudflare.com](https://blog.cloudflare.com/workers-ai-partner-models/)) ## Layer 2: model control — AI Gateway AI Gateway sits between your application and one or more AI providers. It can provide: - Request and response logs. - cost and usage analytics. - retries and fallback. - model/provider switching. - caching. - rate limits. - budgets. - redaction and log controls. - unified model APIs. - governance of AI traffic. This means Brad’s Command Center could use OpenAI for one job, Anthropic for another, and a Workers AI model for a third—while your application goes through one control layer. Cloudflare announced a broader unified inference layer in April 2026, covering Workers AI and numerous external providers. ([blog.cloudflare.com](https://blog.cloudflare.com/ai-platform/)) ## Layer 3: knowledge retrieval — AI Search and Vectorize AI Search, formerly AutoRAG, can: 1. Ingest documents. 2. split them into searchable sections. 3. create embeddings. 4. combine semantic search with traditional keyword search. 5. return relevant source material to an AI model. Vectorize is the underlying general-purpose vector database for custom retrieval and recommendations. Possible applications: - Ask questions about SOPs. - search client contracts. - find the right employee training section. - retrieve similar past sales opportunities. - match a new customer request to a previous solution. - search a family’s private records. - recommend players for a lineup based on comparable scenarios. AI Search supports hybrid semantic and BM25-style keyword retrieval and can be partitioned by customer, language, or agent. ([blog.cloudflare.com](https://blog.cloudflare.com/ai-search-agent-primitive/)) ## Layer 4: state and coordination — Agents and Durable Objects The Agents SDK gives an agent: - A persistent identity. - state. - real-time connections. - tool calls. - scheduled work. - Durable Object storage. - interaction with Workflows and other services. An agent can continue to exist after a user closes the browser. ## Layer 5: durable memory — Agent Memory Agent Memory is intended to retain useful information across conversations and tasks. It can extract structured information, decide what should be remembered, and use namespaces or profiles to isolate users or tenants. As of July 2026 it remains a private-beta service, with pricing not yet generally established. That makes it promising but not yet the place I would store regulated or mission-critical information without explicit contractual assurances. ([developers.cloudflare.com](https://developers.cloudflare.com/agent-memory/)) ## Layer 6: browser use — Browser Run Browser Run lets an agent control a managed Chromium browser. It can: - Navigate sites. - click buttons. - fill forms. - download information. - extract structured data. - generate screenshots or PDFs. - inspect accessibility trees. - run automated tests. - produce Markdown or page snapshots. - allow a human to watch or take over. Cloudflare added live viewing, human-in-the-loop interaction, Chrome DevTools Protocol support, recordings, and greater concurrency in 2026. ([developers.cloudflare.com](https://developers.cloudflare.com/browser-run/quick-actions/)) Browser automation should be used only where: - The website permits it. - APIs are unavailable or inadequate. - credentials are tightly controlled. - actions are logged. - destructive or financial actions require human approval. ## Layer 7: code execution — Sandboxes and Dynamic Workers An agent can generate and test code in a sandbox or execute narrowly permissioned code in a Dynamic Worker. This enables: - Coding agents. - customer-specific formulas. - data-cleaning scripts. - dynamically generated reports. - safe execution of plugins. - tenant-specific business rules. ## Layer 8: files and versioning — R2 and Artifacts R2 stores normal files and objects. Artifacts stores version-controlled file trees through Git. An AI coding agent can: 1. Create an isolated repository. 2. modify source files. 3. run tests in a sandbox. 4. commit changes. 5. present a diff. 6. wait for approval. 7. deploy through Workers. ## Layer 9: durable execution — Workflows Workflows allow an agent to pause for hours or days, retry failures, wait for approval, and continue without losing state. ## Layer 10: private resources — Mesh and Workers VPC An agent may need access to: - A private database. - an internal CRM. - a customer’s private API. - a service running in AWS. - an internal document system. Mesh, Workers VPC, Access, and Tunnel provide the private connectivity and authorization rather than opening those systems to the Internet. ([cloudflare.com](https://www.cloudflare.com/press/press-releases/2026/cloudflare-launches-mesh-to-secure-the-ai-agent-lifecycle/)) ## Layer 11: MCP governance Cloudflare is integrating Model Context Protocol connectivity with: - Access for agent identity. - AI Gateway for visibility and policy. - managed OAuth. - MCP portals. - Code Mode. - discovery of unsanctioned or “shadow” MCP services. This is the agent equivalent of controlling which employee can access which application. ([blog.cloudflare.com](https://blog.cloudflare.com/agents-week-in-review/)) ## Cloudflare is not yet a no-code Zapier replacement The platform gives you the **runtime and infrastructure**, not a complete catalog of finished business automations. You still generally need to define: - What the agent is allowed to do. - How it authenticates to Gmail, Calendar, Drive, HubSpot, or other systems. - What data it stores. - which events trigger it. - how it handles errors. - where human approval is required. - how each client’s data is isolated. Agent Lee is Cloudflare’s own assistant for Cloudflare accounts; it is not a general finished employee for your business. ([developers.cloudflare.com](https://developers.cloudflare.com/agent-lee/)) --- # 16. A full AI-agent example for Outsource Access Consider a **client-onboarding operations agent**. ## What the user experiences 1. A manager creates a new client. 2. The client receives a secure portal and uploads documents. 3. The agent reads the documents and identifies required tasks. 4. It compares the information to Outsource Access SOPs. 5. It creates a proposed onboarding plan. 6. The manager edits or approves the plan. 7. Tasks are assigned to staff. 8. The agent monitors completion and sends reminders. 9. It logs every action. 10. The client sees status without seeing internal notes. 11. At completion, it generates a final onboarding report. ## What Cloudflare does behind the scenes - **Access + Google Workspace:** employee login. - **External identity provider:** client login. - **Worker:** portal and API. - **D1 or PostgreSQL:** client, employee, task, permission, and workflow records. - **R2:** uploaded documents and generated reports. - **AI Search:** searches SOPs and approved client documents. - **AI Gateway:** controls the selected AI model. - **Agents:** maintains the onboarding agent’s state. - **Workflows:** runs the multi-stage onboarding process. - **Queues:** processes uploaded files in the background. - **Browser Run:** uses a third-party portal only where an API does not exist. - **Turnstile/WAF/rate limiting:** protects public forms and authentication. - **DLP:** detects sensitive data leaving through inappropriate channels. - **Logs:** creates an audit record. - **Flagship:** gradually releases the agent to selected teams or clients. - **R2 backups or external replication:** protects against accidental deletion or platform incidents. That is a sophisticated commercial product, and almost all its infrastructure can run through Cloudflare. --- # 17. Documented companies and how they use Cloudflare The following are based on Cloudflare customer stories or direct company statements—not technology-detection guesses. | Company or organization | Documented Cloudflare use | What the user experiences | |---|---|---| | **Canva** | WAF, DDoS protection, Bot Management, Page Shield, CDN, Access, Tunnel, WARP, Browser Isolation, and Workers. | Fast access to Canva, protection from abuse, secure employee access, and edge application logic. | | **Indeed** | WAF and bot protection against fraud; Access/Zero Trust replacing VPN-style access to internal and cloud resources. | Job seekers encounter less fraud; employees connect according to identity, device, and location. | | **Shopify** | Workers for merchant-edge logic, security, personalization, and Shopify’s global marketing environment. | Merchant storefront logic can execute close to shoppers without operating separate regional infrastructure. | | **Caliente.mx** | CDN, Pages, Workers, Stream, and R2; reported substantial cost savings. | Fast betting content, scalable application delivery, video, and storage. | | **VSCO** | Workers AI supporting generative capabilities in VSCO Canvas. | Users receive AI-assisted creative features inside the product. | | **Liveblocks** | Durable Objects, R2, Queues, and KV. | Real-time collaborative application state, attachments, version histories, and background work. | | **Tightknit** | Workers, KV, Queues, and Workflows. | Community automation and workflows without running a conventional server fleet. | | **The Hindu Group** | More than 80 applications deployed using Workers. | Numerous media applications run close to readers on a shared platform. | | **Cinder** | Workers connecting third-party data to no-code trust-and-safety workflows. | Faster fraud and abuse review; Cloudflare reported detection of thousands of fraudulent applications. | | **CrazyGames** | Workers and Images for personalization and image handling. | Game visitors receive optimized imagery and customized experiences. | | **World University Service of Canada** | DNS, WAF, Access, and internal application protection. | Public sites remain protected while staff access internal systems securely. | | **Olive Young** | Bot Management, WAF, and WAN-related services; evaluating additional AI and Workers capabilities. | E-commerce traffic is protected from bots and network disruption. | | **Zerodha** | WAF, DDoS protection, CDN, and DNS. | Faster and more resilient access to financial services, with sharply reduced origin bandwidth. | | **LendingTree** | Workers for CORS, rewriting, inspection, experimentation, routing, TLS, and anti-scraping controls. | Requests are securely routed and transformed while experiments and fraud controls run at the edge. | | **Padlet** | Stream. | Users can upload and play video without Padlet operating its own encoding and global delivery system. | | **PhonePe** | Stream for adaptive video. | Users receive video optimized for their device and connection. | | **TownNews** | Magic Transit. | News infrastructure receives network-level DDoS protection under a predictable commercial structure. | | **Werner Enterprises** | Email Security. | Employees receive fewer malicious phishing and business-email-compromise messages. | | **Bank of Cyprus** | Magic Transit. | Large network attacks can be detected and mitigated within seconds. | | **Legato** | Email Security. | The organization reported a major reduction in weekly phishing incidents. | | **Umbraco** | DDoS protection, HTTPS, and Workers for authentication, deployment, image processing, and request management. | Hosted CMS customers receive secure custom application behavior without separate edge servers. | | **Bumrungrad International Hospital** | Caching, WAF, geographic controls, and Images. | Patients and visitors receive a faster, protected hospital website with optimized media. | | **C&A** | Workers for e-commerce experimentation. | Product and checkout experiences can be tested without waiting for large origin releases. | | **RightBlogger** | Workers AI for image generation. | Users can create images inside a writing and content product. | | **Wikimedia** | Magic Transit. | Wikimedia’s IP network receives large-scale DDoS mitigation. | | **Blibli** | Magic Transit and Network Firewall. | E-commerce network traffic is protected and filtered before reaching internal infrastructure. | ([cloudflare.com](https://www.cloudflare.com/case-studies/canva/)) Cloudflare has also publicly featured relationships or customer logos including Character.AI, Intercom, DoorDash, Discord, Zendesk, Lovable, npm, SiteGPT, Fossil, Visa, Uber, Broadcom, the U.S. Department of Commerce, HubSpot, L’Oréal, Roche, the U.S. Department of Homeland Security, Braze, Carrefour, Delivery Hero, SoFi, Telus, Workday, Colgate-Palmolive, Labcorp, NCR Voyix, Garmin, Genuine Parts, Mars, Telefónica, Anthropic, Asana, Atlassian, Block, CoreWeave, Leonardo.ai, and Stripe. A featured logo confirms a relationship but does **not** by itself establish which exact products are in use. ([cloudflare.com](https://www.cloudflare.com/)) Character.AI, Leonardo.ai, Lexica, and SiteGPT have specifically been identified in Cloudflare materials as R2 users. ([cloudflare.com](https://www.cloudflare.com/press/press-releases/2023/r2-is-the-infrastructure-powering-leading-ai-companies/)) --- # 18. Likely architectures by company type—not claims about specific customers When a company is publicly associated with Cloudflare but has not disclosed exact products, it is safer to reason by use-case rather than pretend to know its configuration. | Company type | Likely useful Cloudflare stack | |---|---| | **E-commerce marketplace** | CDN, Smart Shield, WAF, Bot Management, fraud controls, Waiting Room, Workers, Images, R2, rate limiting, API Shield. | | **AI chat or agent product** | Workers, Durable Objects, AI Gateway, Workers AI or external models, Vectorize, AI Search, R2, Queues, Workflows, Browser Run, Agent Memory. | | **B2B SaaS platform** | Workers for Platforms, Cloudflare for SaaS, custom hostnames, Access, D1/Postgres, R2, Queues, Workflows, WAF, API Shield. | | **Media company** | CDN, Stream, Images, R2, Cache Reserve, Smart Shield, bot/crawler controls, Workers, analytics. | | **Financial-services company** | Magic Transit, Network Firewall, WAF, bot/fraud controls, API Shield, Access, DLP, Email Security, Data Localization, strong key management. | | **Internal operations platform** | Access, Google identity, Workers, D1/Postgres, R2, Workflows, Queues, AI Search, Gateway, DLP, Logs. | | **Consumer family application** | External authentication, Workers, D1/Postgres, R2, Durable Objects, AI Search, AI Gateway, Turnstile, WAF, client-side encryption for highly sensitive files. | --- # 19. Current Cloudflare pricing Cloudflare mixes four pricing models: 1. **Per-domain plan:** Free, Pro, Business, Enterprise. 2. **Per-user:** Zero Trust/SASE. 3. **Usage-based developer services:** requests, CPU, rows, storage, operations, browser time, AI usage, video minutes. 4. **Negotiated enterprise contract:** committed spend across services, support, SLAs, compliance, logging, and custom features. Prices below are published self-service prices as of July 22, 2026. Enterprise contracts can differ substantially. ## Website security and performance plans | Plan | Published price | |---|---:| | Free | $0 | | Pro | $20/month when billed annually; $25 month-to-month | | Business | $200/month when billed annually; $250 month-to-month | | Enterprise/Contract | Custom annual agreement | Free includes authoritative DNS, CDN, SSL, and foundational DDoS and security capabilities. Higher plans add more WAF, performance, certificate, support, analytics, uptime, and configuration capabilities. Business and Enterprise positioning includes stronger SLA and compliance features. ([cloudflare.com](https://www.cloudflare.com/plans/)) ## Zero Trust/SASE | Plan | Published price | |---|---:| | Free | Up to 50 users | | Pay-as-you-go | $7 per user per month | | Enterprise contract | Custom | Retention, support, networking, data controls, and advanced functionality differ across tiers. ([cloudflare.com](https://www.cloudflare.com/plans/)) ## Workers | Dimension | Free | Paid | |---|---:|---:| | Requests | 100,000/day | 10 million/month included; $0.30/additional million | | CPU | 10 ms/invocation | 30 million CPU-ms/month included; $0.02/additional million CPU-ms | | Static assets | Free | Free | | Minimum paid plan | — | $5/month | Cloudflare’s current example calculates a Worker receiving 100 million monthly requests at an average 7 ms of CPU per request at approximately **$45.40 per month**. ([developers.cloudflare.com](https://developers.cloudflare.com/workers/platform/pricing/)) ## Containers The $5 Workers Paid plan includes: - 25 GiB-hours of memory. - 375 vCPU-minutes. - 200 GB-hours of disk. Additional usage is currently priced at: - $0.0000025 per GiB-second of provisioned memory. - $0.000020 per active vCPU-second. - $0.00000007 per GB-second of disk. Container egress is regionally priced, with 1 TB included in North America and Europe and lower included allotments in several other regions. ([developers.cloudflare.com](https://developers.cloudflare.com/containers/pricing/)) ## Durable Objects | Dimension | Paid inclusion and overage | |---|---| | Requests | 1 million/month; then $0.15/million | | Compute duration | 400,000 GB-seconds/month; then $12.50/million GB-seconds | | SQLite rows read | First 25 billion/month; then $0.001/million | | SQLite rows written | First 50 million/month; then $1/million | | SQLite storage | 5 GB-month; then $0.20/GB-month | Long-lived active WebSockets can create meaningful duration charges unless hibernation is used. Cloudflare’s own example shows a busy always-active pattern costing hundreds of dollars, while a hibernating design can cost around $10 for a comparable number of connections. ([developers.cloudflare.com](https://developers.cloudflare.com/durable-objects/platform/pricing/)) ## D1 | Dimension | Free | Paid | |---|---:|---:| | Rows read | 5 million/day | 25 billion/month included; $0.001/million after | | Rows written | 100,000/day | 50 million/month included; $1/million after | | Storage | 5 GB total | 5 GB included; $0.75/GB-month after | ([developers.cloudflare.com](https://developers.cloudflare.com/d1/platform/pricing/)) ## KV | Dimension | Free | Paid | |---|---:|---:| | Reads | 100,000/day | 10 million/month; then $0.50/million | | Writes | 1,000/day | 1 million/month; then $5/million | | Deletes | 1,000/day | 1 million/month; then $5/million | | List operations | 1,000/day | 1 million/month; then $5/million | | Storage | 1 GB | 1 GB; then $0.50/GB-month | KV is cheap for reads and comparatively expensive for frequent writes, reinforcing its intended use as globally distributed read-heavy storage. ([developers.cloudflare.com](https://developers.cloudflare.com/workers/platform/pricing/)) ## R2 | Dimension | Standard | Infrequent Access | |---|---:|---:| | Storage | $0.015/GB-month | $0.01/GB-month | | Class A operations | $4.50/million | $9/million | | Class B operations | $0.36/million | $0.90/million | | Retrieval | None | $0.01/GB | | Internet egress | Free | Free | The Standard free tier includes 10 GB-month of storage, 1 million Class A operations, and 10 million Class B operations monthly. ([developers.cloudflare.com](https://developers.cloudflare.com/r2/pricing/)) ## Queues, Workflows, and related data services - Queues: approximately **$0.40 per million operations** after included usage. - Workflows: uses Workers request/CPU pricing plus storage and step-based dimensions; the paid allocation includes 500,000 steps per month, with $0.80 per additional 100,000 when billing is enabled. - Hyperdrive database queries: 100,000 per day on Free; unlimited queries on Paid, although the external database still has its own charges. - Analytics Engine: approximately $0.25 per million data points written and $1 per million read queries after included allowances. ([cloudflare.com](https://www.cloudflare.com/plans/)) ## Vectorize Workers Paid currently includes: - 50 million queried vector dimensions per month. - 10 million stored vector dimensions. Overages are approximately: - $0.01 per million queried vector dimensions. - $0.05 per 100 million stored vector dimensions. Cloudflare’s published examples range from pennies for experiments to approximately $23.42 for a comparatively large example workload, before model-inference costs. ([developers.cloudflare.com](https://developers.cloudflare.com/vectorize/platform/pricing/)) ## Workers AI Workers AI includes 10,000 neurons per day and charges approximately **$0.011 per 1,000 neurons** beyond the free allocation on the paid plan. Cloudflare also publishes model-specific equivalent token prices because different models consume very different amounts of GPU computation. ([developers.cloudflare.com](https://developers.cloudflare.com/workers-ai/platform/pricing/)) For example, Cloudflare’s July 2026 list showed: - Small Llama models at pennies per million input tokens. - 70-billion-parameter models at roughly $0.293 per million input tokens and $2.253 per million output tokens. - More computationally intensive reasoning models at higher rates. The external-provider cost for models accessed through AI Gateway is separate unless unified billing arrangements apply. ## AI Gateway Core AI Gateway functionality can be used without a separate base charge. Limits differ for retained logs and gateway scale; the Free tier has supported approximately 100,000 logs, while paid configurations can support much more. External model-provider charges remain payable. ([developers.cloudflare.com](https://developers.cloudflare.com/ai-gateway/reference/pricing/)) ## Browser Run Published rates have included: - Free: 10 browser minutes per day and up to three concurrent browsers. - Paid browser time: approximately $0.09 per browser-hour. - Additional concurrency: approximately $2 per concurrent browser allocation. Complex extraction, long-running sessions, recordings, and high concurrency can make browser automation materially more expensive than ordinary Workers. ([cloudflare.com](https://www.cloudflare.com/plans/)) ## Email Service On the Workers Paid plan: - First 3,000 outbound emails per month included. - Approximately $0.35 per additional 1,000 outbound emails. - Inbound routing is unlimited, although Worker processing may incur Worker usage. ([developers.cloudflare.com](https://developers.cloudflare.com/email-service/platform/pricing/)) ## Stream - Storage: $5 per 1,000 minutes of video stored. - Delivery: $1 per 1,000 minutes viewed or delivered. - Ingress and encoding are included. - No separate bandwidth charge is added to delivery. ([developers.cloudflare.com](https://developers.cloudflare.com/stream/pricing/)) ## Other published prices - Smart Shield plus Argo: starting at $5/month. - Load Balancing: starting at $5/month. - APO: $5/month on the Free plan; included with Pro, Business, and Enterprise. - Advanced Certificate Manager: $10/month. - Log Explorer: first 10 GB free, then approximately $1/GB ingested. - Workers Logs: 20 million events/month included on Paid, then $0.60/million. - Workers Logpush: $0.05/million requests. - Workers Builds: $0.005/minute. - Zaraz: $5/million events. - Registrar: at-cost pricing, with some domains starting around $7.85. ([cloudflare.com](https://www.cloudflare.com/plans/)) --- # 20. Example monthly Cloudflare costs These are modeled examples, not named-customer invoices. ## Small internal Outsource Access portal Assumptions: - Fewer than 50 employees. - Free Cloudflare One tier. - Workers Paid. - Fewer than 10 million application requests. - Less than 5 GB D1 data. - Less than 10 GB R2 files. - No major browser, video, or AI use. **Potential Cloudflare infrastructure cost: approximately $5 per month**, plus the domain and any external identity, AI-model, email, monitoring, or engineering costs. That low number is real, but it does not include the cost of correctly designing, building, testing, securing, and operating the software. ## 250 employees using pay-as-you-go Zero Trust 250 × $7 = **$1,750 per month**, before enterprise networking, support, email security, CASB, DLP, or negotiated discounts. ## One tebibyte of standard R2 storage 1,024 GB less 10 GB free = 1,014 billable GB. 1,014 × $0.015 = approximately **$15.21 per month**, before request operations. ## 100 browser-automation hours 100 × $0.09 = approximately **$9**, plus concurrency allocation, Worker usage, and any associated storage. ## 100,000 outbound application emails After 3,000 included: 97,000 ÷ 1,000 × $0.35 = approximately **$33.95**, plus the $5 Workers plan. ## 10,000 stored video minutes 10 × $5 = **$50 per month**. ## One million delivered video minutes 1,000 × $1 = **$1,000**. This illustrates an important point: ordinary Workers, D1, R2, and Vectorize can be extraordinarily inexpensive, while media delivery, browser automation, logs, active real-time objects, enterprise security, and AI can become meaningful. --- # 21. Why Cloudflare can be so inexpensive ## 1. One network is reused for many products DNS, caching, WAF, bot detection, Workers, storage access, Zero Trust, and AI traffic run across a shared network rather than separate product-specific regional infrastructure. ## 2. Much of the marginal work is tiny A Worker that performs five milliseconds of computation or a cache lookup has very low marginal cost. ## 3. Free users fill unused capacity and improve the system Cloudflare has said the free tier helps use otherwise idle network capacity and supplies broad traffic diversity. It also helps Cloudflare see new attacks and deployment patterns earlier. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000016/cloud-20251231.htm)) ## 4. It is a product-led acquisition strategy The free plan removes procurement and sales barriers. A developer can adopt Cloudflare with a domain and eventually bring it into a larger organization. ## 5. Security improves with network scale More traffic gives Cloudflare more signals about bots, exploits, abusive IPs, crawlers, routing failures, and browser behavior. That intelligence can benefit the entire customer base. ## 6. Self-service products have low sales overhead A $5 Worker account does not require a salesperson, solutions architect, lawyer, or procurement process. ## 7. Enterprise customers subsidize significant service expectations Large enterprises pay custom prices for: - Support. - SLAs. - compliance. - log retention. - large-network protection. - dedicated capacity. - advanced bot and fraud controls. - data localization. - contract terms and liability. - account teams. ## The hidden costs Cloudflare’s raw infrastructure may be cheap, but the full system cost can include: - Software development. - security engineering. - database design. - observability. - testing. - backup architecture. - compliance consulting. - identity products. - AI-model fees. - high-volume logs. - browser automation maintenance. - incident response. - enterprise support. For most custom business software, labor and operational discipline—not Worker requests—will be the dominant cost. --- # 22. Major competitors and what they compete with Cloudflare does not have one single competitor. It competes with different companies in different layers. ## A. Hyperscale clouds **AWS, Microsoft Azure, and Google Cloud** They compete with Cloudflare in compute, storage, databases, networking, CDN, security, AI, and enterprise infrastructure. ### Their advantages - Much broader catalog. - Virtual machines and Windows servers. - Kubernetes and mature container orchestration. - Specialized databases. - large-scale GPU clusters and model training. - deep enterprise procurement relationships. - massive partner ecosystems. - regional service choices. - sophisticated data warehouses and analytics. ### Cloudflare’s advantages - Simpler global deployment. - Security and networking integrated into every request. - Workers often have lower operational overhead. - R2 eliminates normal Internet-egress fees. - Zero Trust, CDN, WAF, and application compute share one control plane. - No need to select and manage many regions for most Worker applications. AWS’s catalog includes EC2, Lambda, container platforms, numerous database engines, storage tiers, networking, API Gateway, CloudFront, and security services—far broader than Cloudflare’s conventional cloud catalog. ([aws.amazon.com](https://aws.amazon.com/products/compute/)) ## B. Frontend and web-application platforms ### Vercel Best known for: - Next.js. - frontend deployment. - preview environments. - integrated functions. - AI SDK and AI Cloud positioning. - developer experience. - streaming web applications. Vercel is often smoother for a team deeply committed to Next.js and Vercel’s development workflow. Cloudflare is stronger as a general network, security, storage, Zero Trust, and edge platform. Vercel’s Hobby plan is free and Pro has recently been advertised around $20 per user with included usage credits. ([vercel.com](https://vercel.com/docs)) ### Netlify Best known for: - Git-based web deployment. - deploy previews. - serverless functions. - frontend workflows. - forms and web integrations. - branch-based releases. - rollback. Netlify can be easier for conventional marketing sites and teams centered on its deployment workflow. Cloudflare offers deeper networking, security, storage, and global application primitives. ([netlify.com](https://www.netlify.com/)) ## C. Backend, database, and authentication platforms ### Supabase Supabase supplies: - Full PostgreSQL. - Auth. - object storage. - real-time subscriptions. - Edge Functions. - database administration. - backups and paid point-in-time recovery. Supabase is often more immediately complete as an application backend because PostgreSQL and user authentication are central to the product. Cloudflare is broader in security, networking, edge compute, and AI delivery. ([supabase.com](https://supabase.com/docs/guides/database/overview)) A strong combination is: - Cloudflare Workers, WAF, CDN, R2, AI Gateway, and Queues. - Supabase PostgreSQL and customer authentication. - Hyperdrive between Workers and Supabase. ### Firebase Firebase is strong in: - Mobile applications. - customer authentication. - Firestore. - real-time data. - analytics. - push notifications. - integration with Google Cloud. ### Neon, PlanetScale, Turso, and MongoDB Atlas These compete mainly in database services: - Neon: serverless PostgreSQL. - PlanetScale: managed MySQL/PostgreSQL offerings and database branching/workflows. - Turso: distributed SQLite. - MongoDB Atlas: managed document databases. Cloudflare can connect to them rather than replace them. ## D. CDN, edge, and application-security competitors - Akamai. - Fastly. - Amazon CloudFront/Shield/WAF. - Google Cloud CDN and Cloud Armor. - Azure Front Door. - Imperva. - F5. - Radware. Akamai is especially strong in large-enterprise content delivery and security. Fastly is known for programmable edge delivery and developer control. Cloudflare’s differentiation is the breadth of its unified network, developer, Zero Trust, and AI services. ## E. Zero Trust and SASE competitors - Zscaler. - Netskope. - Palo Alto Prisma Access. - Cisco. - Microsoft Entra and security products. - Fortinet. - Cato Networks. Cloudflare’s differentiator is that employee security, application security, networking, and developer applications run on the same global network. A dedicated SASE vendor may offer deeper maturity in specific enterprise workflows, endpoint ecosystems, or legacy-network integrations. ## F. AI infrastructure competitors - OpenAI. - Anthropic. - AWS Bedrock. - Google Vertex AI. - Azure AI. - Replicate, now owned by Cloudflare. - Together AI. - Fireworks AI. - Modal. - Hugging Face. - Groq. - specialized vector databases such as Pinecone and Weaviate. Cloudflare is not primarily trying to train the largest frontier model. It is positioning itself as the **distributed inference, routing, security, retrieval, and agent-execution layer** around models. ## G. Identity competitors or complements - Auth0. - Okta. - Clerk. - WorkOS. - Supabase Auth. - Firebase Authentication. - AWS Cognito. Cloudflare Access is excellent for workforce and private-application access. It is not yet a complete substitute for every consumer identity and customer-account-management requirement. --- # 23. What Cloudflare cannot—or should not—replace ## Cloudflare is not an office suite It does not replace: - Gmail. - Google Calendar. - Google Docs. - Google Sheets as a collaborative ad hoc surface. - Google Slides. - Google Chat. - Microsoft 365. ## It is not a complete CRM, ERP, or marketing platform Cloudflare gives you the infrastructure to build one. It does not give you a finished HubSpot, Salesforce, NetSuite, or Marketo. ## D1 is not full PostgreSQL D1 lacks parts of the PostgreSQL ecosystem and is not the default answer for every large or highly relational system. ## Workers are not unrestricted servers Some Node.js modules, system calls, binaries, and long-running server assumptions do not work directly. Containers can cover many of these gaps but bring a different runtime and cost model. ([developers.cloudflare.com](https://developers.cloudflare.com/workers/runtime-apis/nodejs/)) ## Cloudflare does not provide conventional VMs or a general Kubernetes service Containers are significant, but Cloudflare still does not offer the same breadth of arbitrary infrastructure as AWS, Azure, or Google Cloud. ## It is not a frontier-model training cloud Workers AI is focused on inference. Training a huge new foundational model requires specialized GPU-cluster infrastructure elsewhere. ## It is not a complete no-code automation platform It provides primitives. Your team still needs to implement connectors, policies, data models, permissions, and user experiences. ## It does not provide compliance automatically Certifications and a BAA are inputs to your compliance program, not substitutes for one. ## It cannot eliminate concentration risk A single provider can have outages. Critical systems need exports, backups, recovery procedures, and where appropriate an alternate delivery or storage path. --- # 24. Security and compliance ## Cloudflare’s own compliance posture Cloudflare reports certifications and authorizations including: - SOC 2 Type II. - ISO 27001. - ISO 27701. - ISO 27018. - PCI DSS. - FedRAMP Moderate authorization for defined services. - Cloud-security attestations and regional frameworks such as C5, ENS, and IRAP. - privacy programs and data-processing agreements. ([cloudflare.com](https://www.cloudflare.com/trust-hub/compliance-resources/soc-2/)) Cloudflare’s SOC 2 Type II report covers the Security, Availability, and Confidentiality criteria and is available under appropriate confidentiality arrangements. Its current scope includes many application, network, Zero Trust, and developer products—including Workers, R2, D1, Durable Objects, Workers AI, AI Gateway, Queues, Vectorize, Stream, Images, and Browser Rendering—but newly launched services may not enter the audit scope until a subsequent audit cycle. ([cloudflare.com](https://www.cloudflare.com/trust-hub/compliance-resources/soc-2/)) This is particularly important for: - Agent Memory. - Artifacts. - Dynamic Workers. - new Email Service features. - experimental agent infrastructure. - any beta product. Never infer that a new product is automatically covered because another Cloudflare product is covered. ## HIPAA There is no universal government-issued “HIPAA certification” for a cloud platform. Cloudflare can support HIPAA safeguards and offers BAAs to eligible Enterprise customers, but you must: 1. Execute the BAA. 2. identify every service that will touch ePHI. 3. confirm that each service is contractually and technically in scope. 4. configure it appropriately. 5. maintain your own HIPAA program. ([cloudflare.com](https://www.cloudflare.com/learning/privacy/what-is-hipaa-compliance/)) ## What you still have to implement ### Identity and authentication - Individual accounts—not shared employee logins. - enforced multifactor authentication. - hardware security keys for administrators where possible. - employee lifecycle processes. - prompt termination of access. - role-based access. - service accounts separate from people. ### Authorization Every API request must verify: - Who is making the request? - Which organization or family do they belong to? - What role do they have? - Are they authorized for this specific record or file? - Is the operation read, edit, approve, export, or delete? Hiding a button in the user interface is not authorization. The backend Worker must enforce it. ### Tenant isolation Every client-facing table should ordinarily include a tenant or organization identifier. The server—not the browser—should add and validate it. For highly sensitive tenants, consider: - Separate databases. - separate R2 buckets or prefixes with separate credentials. - separate encryption keys. - separate Worker deployments. - separate Cloudflare accounts for the most demanding clients. ### Files - Keep R2 buckets private. - prohibit permanent public document URLs. - issue short-lived signed URLs only after authorization. - scan uploaded files. - store file type, owner, tenant, retention, and classification metadata. - log downloads and exports. - support legal hold and deletion policies. - back up important data outside the primary failure domain. ### Secrets - Never place API keys in frontend JavaScript. - use Workers secrets or Secrets Store. - give each environment separate credentials. - rotate credentials. - never give an AI agent a broad administrator token when a narrow operation will do. ### Audit and monitoring Record: - Login. - failed login. - permission changes. - document views and downloads. - exports. - AI access to sensitive information. - agent tool calls. - administrator actions. - deleted records. - changes to client ownership. - authentication and API errors. ### Backups and disaster recovery R2 durability and D1 replication do not replace a tested backup plan. Maintain: - Scheduled database exports. - versioned or replicated files. - infrastructure-as-code. - documented recovery steps. - recovery-time and recovery-point objectives. - periodic restore tests. ### AI-specific security Treat an AI agent as a potentially confused junior employee with access to tools. Implement: - Tool-level scopes. - tenant-level data isolation. - egress allowlists. - maximum spend and action limits. - prompt-injection defenses. - human approval for money movement, account changes, sending sensitive content, deletion, and irreversible actions. - complete action logs. - separate read and write permissions. - no direct exposure of raw credentials. - content filtering where appropriate. - model-provider contractual review. Cloudflare’s Dynamic Worker binding model, Access, AI Gateway, Mesh, and MCP controls help implement these boundaries, but you must design the policies. ([developers.cloudflare.com](https://developers.cloudflare.com/dynamic-workers/usage/bindings/)) --- # 25. Is R2 an appropriate secure PDF environment? Yes, when implemented properly. A secure document portal can use: - **R2:** encrypted file storage. - **D1/PostgreSQL:** document metadata and permissions. - **Access:** employee login. - **External identity:** client login. - **Worker:** authorization and file delivery. - **DLP:** sensitive-information controls. - **WAF and rate limits:** attack protection. - **Logs:** audit history. - **Queues:** malware scanning, classification, OCR where appropriate, and AI indexing. - **AI Search:** permission-aware document retrieval. However, Google Drive can also securely hold PDFs. The reason to introduce R2 should be that you need a **purpose-built application experience**, stricter application-level permissions, tenant-specific portals, automated workflows, or controlled AI access—not merely because a file is a PDF. For Outsource Access, a sensible transition is: 1. Keep collaborative working documents in Google Drive. 2. Put application attachments and client-portal files into R2. 3. store links to existing Drive items during the first migration. 4. move additional files only when the portal has equivalent permissions, retention, search, backup, and audit behavior. --- # 26. Reliability, incidents, and controversies Cloudflare is critical Internet infrastructure, but it is not infallible. ## Cloudbleed: 2017 A parser bug caused fragments of memory from some requests to be exposed in unrelated responses. Cloudflare worked with search engines and other organizations to remove cached leaked information and published a detailed explanation. The incident remains an important example of the risk created when one reverse proxy handles enormous amounts of traffic. ([blog.cloudflare.com](https://blog.cloudflare.com/incident-report-on-memory-leak-caused-by-cloudflare-parser-bug/)) ## Significant 2025 outages - A June 12, 2025 incident involving a shared storage dependency affected numerous Cloudflare products for approximately two and a half hours. - A November 18, 2025 incident was triggered by an oversized bot-management configuration file. - A December 5, 2025 incident lasted approximately 25 minutes and affected a substantial share of traffic. Cloudflare subsequently announced resilience and dependency-reduction work. ([blog.cloudflare.com](https://blog.cloudflare.com/cloudflare-service-outage-june-12-2025/)) ## 2026 issues Cloudflare reported and patched an ACME/WAF logic issue in early 2026 and said it found no evidence of exploitation. A February 2026 BYOIP incident also affected parts of the network. ([blog.cloudflare.com](https://blog.cloudflare.com/acme-path-vulnerability/)) ## Content-policy controversies Because Cloudflare is an infrastructure intermediary, it has repeatedly faced questions about whether and when it should stop serving harmful or extremist sites. Decisions involving The Daily Stormer, 8chan, and Kiwi Farms became prominent examples. Cloudflare’s filings acknowledge that either serving or terminating controversial customers can cause reputational, political, regulatory, and commercial risk. ([sec.gov](https://www.sec.gov/Archives/edgar/data/1477333/000147733326000016/cloud-20251231.htm)) **Lesson for you:** Cloudflare should be a major foundation, but your databases and files must remain exportable and your recovery plan should not depend on Cloudflare being continuously available. --- # 27. Recommended Outsource Access architecture ## Identity ### Employees Use: - Google Workspace as identity provider. - Cloudflare Access for every internal application. - Workspace groups such as `oa-admins`, `oa-managers`, `oa-client-A-team`. - MFA and device posture. - separate service tokens for automations. ### Clients Use a customer-identity provider supporting: - Invitations. - passwordless login or strong MFA. - account recovery. - organization membership. - client administrators. - audit events. - optional enterprise SSO. Access can protect client portals, but a dedicated customer-identity product may produce a better external account experience. ## Frontend Use Workers or Pages with: - React, Astro, or another familiar framework. - reusable dashboards. - responsive mobile layouts. - separate internal and client navigation. - feature flags through Flagship. ## APIs Use Workers for: - Authorization. - validation. - workflow commands. - document delivery. - integrations. - AI calls. - webhook handling. ## Database Start with D1 when: - The app is modest. - The relational model is clear. - Reporting is not extreme. - You value simplicity and low cost. Use PostgreSQL through Hyperdrive when: - You need complex reporting. - You expect a large product organization. - You need extensive relational tooling. - You need mature row-level security or extensions. - You want Supabase’s authentication and administration. - Portability is a major concern. Build the application through a data-access layer so that moving from D1 to PostgreSQL does not require rewriting the entire user interface. ## Files Use private R2 buckets with prefixes such as: - `clients/{tenant-id}/documents/` - `clients/{tenant-id}/reports/` - `employees/{employee-id}/private/` - `shared/training/` - `audit-exports/` Never trust a client-supplied tenant identifier without matching it to the authenticated account. ## Background work - Queues for individual jobs. - Workflows for business processes. - scheduled Workers for recurring synchronization. - Durable Objects for real-time status. ## AI - AI Gateway as the universal model gateway. - Workers AI for inexpensive local models. - OpenAI, Anthropic, or another provider for model-specific strengths. - AI Search for approved SOPs and client data. - separate retrieval indexes or namespaces by tenant. - human approval for actions. - no unreviewed write access to Gmail, calendars, banking, HR, or client systems. ## Security - WAF. - Turnstile on public forms. - rate limits. - API Shield for partner APIs. - DLP for sensitive data. - private origins through Tunnel. - no public database. - strict logs and alerts. - infrastructure managed through Terraform. - separate development, staging, and production accounts or environments. --- # 28. Application blueprint: replacing spreadsheets A Google Sheet often combines five separate things: 1. Database. 2. user interface. 3. permissions. 4. workflow. 5. reporting. A proper application separates them. ## Example: staff/client assignment system ### Database tables - Employees. - clients. - employee-client assignments. - roles. - skills. - availability. - tasks. - statuses. - notes. - approvals. - audit events. ### User experience An employee sees: - Their assigned clients. - today’s tasks. - overdue items. - SOP links. - messages requiring attention. - permitted client documents. A manager sees: - Capacity by employee. - unassigned work. - late tasks. - quality issues. - training gaps. - client risk. - approval queues. A client sees: - Deliverables. - status. - approved documents. - requests. - reports. - communication history intended for clients. ### Cloudflare components - Access/Google identity. - Workers. - D1/Postgres. - R2. - Workflows. - Queues. - Durable Objects for live updates. - AI Search. - AI Gateway. - WAF and logs. A spreadsheet can remain available as an export, but it should no longer be the authoritative workflow engine. --- # 29. Application blueprint: coach lineup tool ## Data - Teams. - coaches. - players. - positions. - availability. - opponent. - formations. - playing-time rules. - player combinations. - injuries. - previous lineups. - performance indicators. ## User experience A coach: 1. Selects the upcoming game. 2. marks unavailable players. 3. sets constraints. 4. asks the tool for three lineup options. 5. sees why each lineup was selected. 6. drags players to adjust. 7. shares the final lineup. 8. records what happened. ## Cloudflare stack - Workers/Pages: interface. - D1: team and lineup records. - Durable Objects: simultaneous coach editing. - Workflows: reminders and lineup approval. - R2: player photos and exports. - AI Gateway/Workers AI: natural-language explanations. - optional optimization code in a container or sandbox. - Turnstile and WAF: public registration protection. This product can run almost entirely on Cloudflare. --- # 30. Application blueprint: Family Hub ## Potential features - Household calendar. - emergency contacts. - school information. - medical contacts. - insurance policies. - household inventory. - warranties. - estate information. - travel plans. - routines. - family tasks. - private AI assistant. - document search. - emergency-access mode. ## Architecture - External family authentication with MFA. - each family represented as a tenant. - D1/Postgres for structured information. - R2 for private documents. - Durable Object for live family updates. - AI Search with one isolated index or namespace per family. - AI Gateway for model control. - client-side encryption for the most sensitive documents. - recovery keys and an emergency-access policy. - audited document viewing and sharing. - no model training on family data. This product deserves a higher privacy standard than an ordinary social application, even when the information is not legally regulated. --- # 31. Application blueprint: Brad’s Command Center ## Data sources - Gmail. - Google Calendar. - Google Contacts. - tasks. - notes. - personal CRM relationships. - projects. - sales opportunities. - follow-ups. - habits. - selected documents. - communication summaries. ## Recommended flow 1. Google OAuth grants narrowly scoped access. 2. Workflows synchronize incremental changes. 3. Queues process messages and calendar events. 4. D1 or PostgreSQL stores normalized metadata. 5. R2 stores only necessary attachments or generated artifacts. 6. AI Gateway routes summarization and planning to the selected model. 7. AI Search retrieves relevant relationships, projects, and notes. 8. Durable Objects push updated dashboard status in real time. 9. A personal agent proposes tasks and follow-ups. 10. Sending, deletion, and calendar changes require explicit approval. ## Important design principle Do not copy your entire Gmail account into a new database merely because it is possible. Store: - Message identifiers. - useful extracted metadata. - task relationships. - summaries where appropriate. - minimal cached content. - a link back to the source. Fetch the original message when needed. This reduces security exposure, storage, synchronization complexity, and stale duplicate data. --- # 32. Building software for clients For a multi-client software business, consider this progression. ## Stage 1: one shared application with tenant isolation - One frontend. - one API. - one database with tenant identifiers. - separate permissions. - separate R2 prefixes. - per-tenant AI indexes. This is the simplest and most economical. ## Stage 2: client-branded portals Use Cloudflare for SaaS: - `portal.client-one.com` - `ops.client-two.com` - automatic custom certificates. - tenant-specific branding and routing. ## Stage 3: tenant-specific code or rules Use: - Workers for Platforms. - Dynamic Workers. - tenant-specific workflows. - tenant-specific feature flags. Each client can have special logic without requiring a completely separate conventional server deployment. ## Stage 4: stronger isolation for sensitive clients For regulated or large clients: - Separate database. - separate R2 bucket. - separate encryption keys. - separate Cloudflare account or enterprise zone. - separate logs. - contractual data residency. - stricter deployment approvals. --- # 33. HIPAA reference architecture For an application that will store or transmit ePHI: ## Contract layer - Cloudflare Enterprise agreement. - BAA. - current service-scope confirmation. - DPA. - subcontractor review. - breach-notification terms. - retention and deletion obligations. - confirmation of any external AI provider’s BAA. ## Technical layer - Access or customer identity with MFA. - least-privilege roles. - private Workers APIs. - private R2. - encrypted database. - short-lived file access. - immutable or strongly protected audit logs. - DLP. - secrets management. - tenant isolation. - verified backups. - incident alerts. - no ePHI in URLs, analytics labels, or general logs. - no ePHI sent to uncovered beta services. - human approval before high-impact agent actions. ## Organizational layer - Security risk assessment. - HIPAA policies. - workforce training. - access-review cadence. - termination procedures. - vendor inventory. - incident-response plan. - disaster-recovery tests. - sanctions policy. - documented minimum-necessary access. ## Important caution I would not make Agent Memory, Artifacts, or another private-beta service the primary repository for ePHI unless Cloudflare expressly adds it to the BAA and applicable audit scope. This architecture guidance is not itself a legal or compliance opinion; a qualified HIPAA professional should validate the final data flow and contracts. --- # 34. SOC 2 for Outsource Access Cloudflare’s SOC 2 report helps answer: > “Does Cloudflare operate its covered systems with audited controls?” It does not answer: > “Does Outsource Access operate its application securely?” To pursue your own SOC 2 readiness, you will need evidence around: - Employee onboarding and termination. - MFA. - access reviews. - code review. - production deployment. - change management. - incident response. - risk management. - vendor management. - backup tests. - vulnerability management. - monitoring. - security training. - business continuity. - customer-data deletion. - AI usage and model-provider controls. Cloudflare can provide technical mechanisms and vendor evidence for many of these, but your policies and actual behavior generate your SOC 2 evidence. --- # 35. Risks of vibe coding on Cloudflare Vibe coding can make application creation dramatically faster. The main danger is that an interface that appears to work can conceal serious backend problems. Common failures include: - Trusting a tenant ID sent by the browser. - storing secrets in frontend code. - no authorization on API routes. - public R2 buckets. - unrestricted database queries. - missing rate limits. - no input validation. - direct AI access to administrative tools. - no backups. - no schema migration plan. - no tests. - one Cloudflare API token used across all clients. - production and testing using the same data. - logging sensitive records. - allowing an AI agent to follow instructions from untrusted documents. - assuming a successful login means the user can access every record. ## Minimum production standard Every serious application should have: - A written data model. - a written permission matrix. - automated tests for tenant separation. - code review by someone who understands security. - infrastructure-as-code. - separate staging and production. - least-privilege credentials. - backup and restore tests. - monitoring. - an incident plan. - dependency scanning. - explicit AI tool permissions. - a human approval layer for irreversible actions. AI should accelerate this work, not eliminate the controls. --- # 36. Cloudflare’s marketing and sales lessons for Outsource Access ## 1. Start with a free or extremely low-friction product Cloudflare’s free plan lets a user experience genuine value before sales involvement. For Outsource Access, this could become: - Free workflow assessment. - free client-portal prototype. - limited AI knowledge assistant. - spreadsheet-to-application diagnostic. - security and access review. ## 2. Land with one problem and expand Cloudflare often lands with DNS or CDN and later expands into security, networking, and compute. Outsource Access can land with: - A client dashboard. - a single workflow. - a document portal. - an onboarding system. - a reporting application. Then expand into: - AI search. - task automation. - approvals. - sales support. - employee performance. - client analytics. - additional departments. ## 3. Sell a platform outcome rather than individual technology Cloudflare increasingly sells the “connectivity cloud,” not 60 isolated tools. Outsource Access can sell: > A secure client operations platform that connects your people, documents, workflows, and AI—not a collection of automations. ## 4. Publish outcome-based case studies Cloudflare case studies emphasize: - Bandwidth reduction. - incident reduction. - faster page delivery. - fewer phishing messages. - deployment speed. - infrastructure savings. Your case studies should quantify: - Hours saved. - spreadsheet errors eliminated. - onboarding time reduced. - follow-up compliance improved. - response time improved. - number of systems consolidated. - client visibility increased. - employee capacity created. ## 5. Use a reusable core Cloudflare built a common network and turns it into multiple products. Outsource Access should build a common software core containing: - Authentication. - tenant isolation. - client profiles. - documents. - tasks. - approvals. - notifications. - audit logs. - AI gateway. - reporting. - integrations. Then package vertical solutions on top of it. ## 6. Combine product-led growth with managed service The software can be low-cost because Cloudflare is inexpensive. The valuable commercial layer is: - Configuration. - workflow design. - integration. - data migration. - agent design. - security. - training. - ongoing management. That supports recurring managed-service revenue rather than a one-time software project. ## 7. Be transparent about failures Cloudflare’s detailed outage and security postmortems have been an important trust-building tactic, even when the underlying events were serious. For client software, adopt: - Public or client-visible status. - clear incident summaries. - root-cause analysis. - corrective-action tracking. - no vague excuses. --- # 37. Recommended platform standard for Outsource Access I would define the following as the default OA application stack: | Layer | Default | |---|---| | Employee identity | Google Workspace through Cloudflare Access | | Customer identity | Clerk, Auth0, WorkOS, or Supabase Auth | | Frontend | Workers/Pages with Astro or React | | API | TypeScript Workers | | Simple database | D1 | | Complex database | PostgreSQL through Hyperdrive | | Files | Private R2 | | Real-time state | Durable Objects | | Background work | Queues | | Business processes | Workflows | | AI control | AI Gateway | | Models | Workers AI plus selected external providers | | Knowledge retrieval | AI Search/Vectorize | | Browser automation | Browser Run, with approval and restricted credentials | | Heavy code | Containers/Sandbox SDK | | Feature releases | Flagship | | Secrets | Secrets Store/Worker secrets | | Public-form protection | Turnstile, WAF, and rate limiting | | Internal connectivity | Tunnel, Access, Workers VPC | | Configuration | Terraform | | Logs | Workers Logs/Logpush plus retained audit storage | | Backups | Scheduled exports to a separate recovery destination | This is not “Cloudflare only.” It is **Cloudflare first, with specialized providers where they are clearly stronger**. --- # 38. Suggested implementation sequence ## Foundation - Establish production, staging, and development environments. - configure Google Workspace identity through Access. - create Terraform-controlled Cloudflare configuration. - define tenant, role, and permission standards. - define logging, backups, and incident procedures. - establish AI-provider and data-classification policies. ## First product: internal operations hub Build: - Client directory. - employee-client assignments. - task and approval system. - SOP access. - document links. - basic reporting. Avoid AI initially except for low-risk summarization. Establish correct data and permission models first. ## Document and knowledge layer Add: - Private R2. - document metadata. - permission-aware downloads. - AI Search. - controlled document ingestion. - citations in generated answers. - audit history. ## Workflow automation Add: - Queues. - Workflows. - email/calendar integrations. - reminders. - manager approval. - exception handling. ## Reusable client portal Add: - Customer identity. - tenant branding. - custom domains. - client-facing status. - secure uploads. - reports. - request management. ## Agent layer Add: - AI Gateway. - agent state. - narrowly scoped tools. - browser automation only where needed. - human approvals. - cost controls. - complete action logging. ## External software products Reuse the same foundation for: - Coach lineup product. - Family Hub. - Brad’s Command Center. - client-specific workflow products. - industry-specific portals. --- # 39. Cloudflare’s 2026 direction and most important recent news Cloudflare’s latest strategic direction is the **agentic cloud**: - April 2026: unified model access, Agent Lee, Mesh, Workflows improvements, Email Service, Flagship, Dynamic Workers, Sandboxes, Artifacts, and related agent infrastructure. - May 2026: announced approximately 1,100 workforce reductions and an AI-oriented organizational restructuring. - June 2026: Agent Memory private beta; VoidZero and Ensemble AI teams joined Cloudflare. - July 2026: Workers Cache, expanded AI-crawler controls, Attribution Business Insights, content monetization efforts, an OpenAI content-quality/freshness pilot, and Precursor bot validation. - September 15, 2026: Cloudflare has announced planned default AI-crawler policy changes for certain new/free sites, subject to customer configuration. ([cloudflare.com](https://www.cloudflare.com/press/press-releases/2026/cloudflare-expands-its-agent-cloud-to-power-the-next-generation-of-agents/)) Cloudflare is attempting to control not just how AI applications run, but also: - How agents identify themselves. - how websites permit or deny agent usage. - how content owners measure AI crawler value. - how AI companies pay for content use. - how agents securely access private systems. - how agent-generated code is isolated. - how models and providers are routed. - how agents remember information. - how humans approve their actions. That is considerably more ambitious than being a CDN or serverless host. --- # 40. The major strategic risks to Cloudflare ## Technology breadth may outpace maturity The product catalog is expanding faster than most organizations can evaluate it. Some new services remain beta, private beta, or only recently generally available. ## AI economics Inference and GPU services can pressure gross margins. Cloudflare must prove that agent workloads generate sufficient revenue and do not merely increase infrastructure costs. ## Developer competition Vercel, AWS, Supabase, Netlify, and specialized database or AI providers often offer a more focused or mature experience for a particular workload. ## Enterprise execution Cloudflare must sell larger contracts, support complex organizations, and integrate many acquired capabilities without losing its simplicity. ## Reliability concentration A shared global control plane creates efficiency but can turn a dependency or configuration error into a wide product outage. ## Regulatory and content pressure Its position between sites, users, crawlers, governments, and AI companies creates ongoing policy conflicts. ## Founder control Founder control supports long-term decisions but reduces the ability of outside shareholders to change direction. ## Valuation The stock price assumes substantial future success. Even strong growth can be punished if it falls below those expectations. --- # 41. A beginner’s Cloudflare glossary | Term | Plain-English meaning | |---|---| | **DNS** | The system that converts a name such as `example.com` into a network address. | | **Reverse proxy** | A service that receives visitors before forwarding requests to the real application. | | **Origin** | The underlying server or service from which Cloudflare retrieves content. | | **CDN** | A worldwide cache that puts copies of content close to users. | | **Edge** | Infrastructure near users rather than in one central data center. | | **Anycast** | Advertising the same IP address from many locations so traffic reaches a nearby location. | | **TLS/SSL** | Encryption used for HTTPS. | | **WAF** | A firewall that understands web requests and blocks application attacks. | | **DDoS** | An attack that overwhelms a service with massive traffic. | | **Bot** | Automated software making requests; it can be helpful, neutral, or malicious. | | **Serverless** | Run code without provisioning or maintaining a conventional server. | | **Worker** | Cloudflare’s globally distributed serverless code runtime. | | **Container** | A packaged application with its runtime, libraries, and operating-system environment. | | **Database** | Structured, queryable application records. | | **SQL** | A language for querying relational databases. | | **Object storage** | Storage for files and blobs rather than rows and columns. | | **Key-value storage** | Information stored and retrieved by a unique key. | | **Vector database** | Stores numeric representations of meaning for AI similarity search. | | **Queue** | A durable line of jobs waiting to be processed. | | **Workflow** | A multi-step process that can wait, retry, and retain progress. | | **API** | A structured way for one software system to communicate with another. | | **Webhook** | An event message one system sends to another when something happens. | | **Tenant** | One customer or organization inside a shared software application. | | **SSO** | One identity used to log in to multiple applications. | | **Identity provider** | The system that verifies users, such as Google Workspace or Okta. | | **Zero Trust** | Verify every access request rather than trusting someone because they are “inside the network.” | | **RAG** | Retrieving relevant documents and providing them to an AI model before it answers. | | **Embedding** | A numeric representation of text, an image, or another object’s meaning. | | **MCP** | A standard through which AI agents discover and call tools or data sources. | | **Egress** | Data leaving a cloud provider, often a separate cost elsewhere. | | **Infrastructure as code** | Version-controlled configuration that creates and manages cloud resources reproducibly. | | **BAA** | A HIPAA-required business-associate agreement for vendors handling protected health information. | | **SOC 2 Type II** | An independent audit of whether defined controls operated over a period of time. | --- # Final assessment Cloudflare is now one of the broadest software infrastructure platforms available. Its unique strength is not that it has the world’s best version of every individual database, authentication product, model, or server. Its strength is that it combines: - Global networking. - application security. - employee security. - serverless compute. - stateful compute. - conventional containers. - object storage. - SQL. - asynchronous workflows. - real-time systems. - media. - browser execution. - AI inference. - AI retrieval. - agent state and memory. - code isolation. - multitenant application infrastructure. - agent and MCP security. For Outsource Access, I would adopt Cloudflare as the **secure application, workflow, and AI foundation**, while keeping Google Workspace as the communication, document-collaboration, and employee-identity system. The first strategic product should be a reusable **Outsource Access Operations Platform** with: 1. Google-based employee login. 2. client and employee records. 3. tasks and approvals. 4. private R2 documents. 5. client dashboards. 6. audit logs. 7. workflows. 8. permission-aware AI search. 9. AI Gateway. 10. a reusable tenant and role architecture. Once that foundation is correct, the coach lineup tool, Family Hub, Brad’s Command Center, and client-specific portals become different configurations of the same core platform rather than four unrelated software projects. The largest mistake would be treating Cloudflare’s low infrastructure price as evidence that secure software is effortless. The infrastructure can be extremely inexpensive. Correct authorization, tenant isolation, data governance, backups, AI permissions, compliance, and product design remain the real work. --- # Turn 3 · User **Category:** Publishing request **Title:** Request to publish the research as an interactive mobile-friendly interface Can you please publish this to interactive engaging easy to read Webb type interface that’s mobile friendly --- # Turn 4 · Assistant **Category:** Artifact delivery **Title:** Delivery of the original interactive Cloudflare Expert Guide # Cloudflare Expert Guide — interactive web edition I converted the research into a responsive web application with: - Searchable exploration of all **128 Cloudflare capabilities** - Interactive company history, funding, acquisition, leadership, employee, revenue, and stock sections - An architecture builder for the Outsource Access hub, client portals, Family Hub, lineup application, Brad’s Command Center, and HIPAA-oriented systems - A visual AI-agent stack explaining models, memory, tools, browsers, workflows, and security guardrails - Filterable customer examples showing which Cloudflare products companies use - A working monthly-cost estimator with several application scenarios - Competitor comparisons for Cloudflare, AWS, Azure, Google Cloud, Vercel, Netlify, Supabase, Firebase, Akamai, Fastly, and others - HIPAA, SOC 2, document-security, tenant-isolation, and AI-agent security guidance - A persistent production-readiness checklist - Dark mode, global search, expandable details, mobile navigation, print/PDF support, and offline-installable web-app metadata ## Open the interactive guide [**Launch the interactive Cloudflare Expert Guide**](sandbox:/mnt/data/cloudflare-expert-guide-published/cloudflare-expert-guide.html) This is the self-contained edition. It can be opened directly without installing anything or running a server. ## Publish it publicly [**Download the Cloudflare drag-and-drop website package**](sandbox:/mnt/data/Cloudflare-Expert-Guide-Site.zip) The ZIP contains `index.html` and all deployment assets at its root. It is prepared for Cloudflare Drop or Cloudflare Pages Direct Upload. Cloudflare Drop accepts a static-site folder or ZIP, immediately provides a temporary `workers.dev` address, and allows the deployment to be claimed within 60 minutes so it remains online. ([cloudflare.com](https://cloudflare.com/drop)) Cloudflare Pages also supports dashboard-based drag-and-drop deployment of a ZIP or folder, resulting in a public `pages.dev` site. ([developers.cloudflare.com](https://developers.cloudflare.com/pages/get-started/direct-upload/?utm_source=chatgpt.com)) ## Full editable project [**Download the complete source and deployment project**](sandbox:/mnt/data/Cloudflare-Expert-Guide-Source.zip) It includes: - Editable HTML, CSS, JavaScript, and research data - Cloudflare Workers Static Assets configuration - Wrangler deployment commands - Installable web-app assets - A deterministic build script - The drag-and-drop deployment ZIP - Desktop and mobile previews - Deployment documentation [**Read the publishing instructions**](sandbox:/mnt/data/cloudflare-expert-guide-published/PUBLISH.md) ## Preview [Desktop interface preview](sandbox:/mnt/data/cloudflare-expert-guide-published/preview/desktop.png) [Mobile interface preview](sandbox:/mnt/data/cloudflare-expert-guide-published/preview/mobile.png) The finished interface was tested at **320, 360, 390, 768, 1,024, 1,440, and 1,920-pixel widths**. The browser tests covered search, filters, capability dialogs, the application-stack builder, pricing calculations, competitor views, HIPAA guidance, the security checklist, roadmap navigation, dark mode, and mobile menus. It produced no JavaScript errors and no page-level horizontal overflow. I could not create the final public Cloudflare URL from this environment because outbound browser uploads are administratively blocked and no Cloudflare account is connected. The drag-and-drop ZIP is ready for the final upload and claim step. --- # Turn 5 · User **Category:** Fidelity question **Title:** Question about whether the dashboard contains the exact original research Okay this interactive platform you built is absolutely unbelievable. I want to understand: the logic is above. You gave me just written output of your research and analysis in plain text form. Is this interactive tool that you made, have every single bit of the exact same text and content in the interactive dashboard tool or is it an entirely different format with a different objective? Please help me understand how you translated the research work then to this insanely interactive guide just from my one simple prompt --- # Turn 6 · Assistant **Category:** Design rationale **Title:** Explanation of how the research was translated into the interactive guide ## The honest answer **No—the interactive guide does not contain every word of the original research response verbatim.** It also is **not an unrelated product with a different objective**. It is a **structured, edited, interactive adaptation of the same research**. Think of the relationship this way: - The original answer is the **comprehensive written research report**. - The interactive guide is the **visual learning, exploration, and decision-making edition** of that report. - The source facts, central conclusions, product taxonomy, examples, and recommendations are substantially the same. - The wording, organization, level of detail, and way you encounter the information are intentionally different. The interface prioritizes comprehension and navigation over preserving every sentence. --- # What was preserved exactly or nearly exactly The strongest one-to-one translation is the Cloudflare product catalog. The application contains structured records for: - **All 128 named Cloudflare platform surfaces** - **13 major historical milestones** - **10 annual revenue data points** - **7 employee-count data points** - **6 funding events** - **7 current leadership profiles** - **16 acquisitions or acquired teams** - **26 documented customer examples** - **6 application architecture blueprints** - **12 competitor profiles** - **23 production-security checklist controls** - **4 interactive pricing scenarios** - **22 primary research-source links** For each of the 128 capabilities, I created: 1. The product name. 2. Its Cloudflare category. 3. A plain-English explanation. 4. A concrete user-experience example. So the original product catalog was not reduced to a handful of highlights. The full taxonomy was converted into searchable, filterable records. --- # What was condensed The long-form report contains substantially more narrative explanation. For example, the original report explains Cloudflare’s evolution through multiple paragraphs about: - Project Honey Pot. - Cloudflare’s founding. - its launch. - Universal SSL. - Workers. - Zero Trust. - R2. - AI infrastructure. - the agentic-cloud strategy. In the interactive interface, that narrative becomes a **13-stage timeline**. The timeline preserves the sequence and meaning, but each stage is summarized into a short explanation that works well on a phone. The same compression happened in several areas: | Original report | Interactive translation | |---|---| | Long corporate-history narrative | Visual timeline and milestone cards | | Revenue, employee, and customer-growth tables | Charts, counters, and market snapshots | | Full investment discussion | Expandable funding history | | Detailed leadership narrative | Leadership directory and role summaries | | 128-product tables | Searchable capability explorer | | Long technical explanations | Plain-English capability cards and examples | | Multiple application architecture sections | Interactive application-stack builder | | AI-agent stack explanation | Clickable layered agent diagram | | Long customer table | Searchable and filterable customer cards | | Large pricing section | Interactive estimator and pricing presets | | Competitor narrative | Comparison lenses and provider cards | | HIPAA and SOC 2 discussion | Compliance views and security checklist | | Implementation recommendations | Seven-stage Outsource Access roadmap | | Technical glossary | Searchable vocabulary translator | The original was written to be read sequentially. The guide was designed to be explored nonlinearly. --- # What is not fully reproduced in the guide Several parts of the original response are present only in shortened form. ## 1. Not every pricing detail is inside the estimator The written report includes pricing commentary for many individual products, including: - Workers. - Containers. - Durable Objects. - D1. - KV. - R2. - Queues. - Workflows. - Vectorize. - Workers AI. - AI Gateway. - Browser Run. - Email Service. - Stream. - Logs. - Zaraz. - certificates. - load balancing. - Smart Shield. - Registrar. The interactive estimator models a more manageable subset: - Workers requests and CPU. - D1 storage. - R2 storage. - Zero Trust seats. - Browser automation. - outbound email. - Stream video. - a manually entered AI budget. That was deliberate. An estimator with every Cloudflare billing dimension would become a complicated procurement calculator rather than an approachable learning tool. ## 2. Customer stories are shortened The original response explains some customer implementations in more detail. The interactive guide turns them into concise cards containing: - Company. - industry. - documented products. - what the products accomplish. - what the end user experiences. It does not reproduce every statistic, quotation, or case-study paragraph. ## 3. Citations are consolidated The report connects many factual claims directly to individual sources. The web guide has a research methodology section and **22 source destinations**, but it does not display a citation after every sentence. That makes the interface cleaner, although it makes the prose report better for rigorous claim-by-claim verification. ## 4. Some corporate nuance is summarized The written report gives more explanation about: - Institutional ownership. - dual-class voting. - stock valuation. - margin structure. - founder control. - organizational changes. - private equity versus venture investment. - reliability incidents. - content-policy controversies. The interface keeps the major findings, numbers, and headlines but does not reproduce every paragraph of analysis. ## 5. Security material is action-oriented rather than exhaustive The full report discusses HIPAA, SOC 2, tenant isolation, file security, AI-agent security, secrets, auditing, backups, and authorization at length. The interface converts those ideas into: - A shared-responsibility model. - HIPAA and SOC 2 views. - an AI-agent security view. - a private-file model. - a 23-control interactive checklist. That is better for implementation, but the original report contains more explanatory nuance. --- # It has a related but somewhat different objective The objectives overlap, but they are not identical. ## Objective of the original report The report was intended to answer: > “What is Cloudflare, how did it evolve, what can it do, who uses it, how much does it cost, how secure is it, and how should Outsource Access use it?” It is an analytical reference document. ## Objective of the interactive guide The guide was intended to answer: > “How can Brad learn, explore, compare, model, and apply this information without reading one enormous document from beginning to end?” It is a learning and decision-support interface. That difference explains why the guide contains things that were not literal sections of the report: - A global search engine. - filtering by capability category. - a product architecture builder. - pricing sliders. - competitor lenses. - persistent checklist completion. - mobile navigation. - dark mode. - expandable dialogs. - an AI-agent flow. - an implementation roadmap selector. Those interactions are **derived from the report’s logic**, rather than copied from its sentences. --- # How one prompt became the interactive guide Your publishing prompt was short: > “Publish this to an interactive, engaging, easy-to-read web-type interface that’s mobile friendly.” But the word **“this”** referred to the entire research report immediately above it. The earlier report was still part of the conversation context. Your second prompt was therefore not starting from zero—it was a transformation instruction applied to a large body of existing source material. I effectively interpreted the request as: > Take the complete Cloudflare report already in the conversation, reorganize it into an interactive information architecture, make it understandable to a non-developer, optimize it for mobile, and package it as a publishable website. Here is how that transformation worked. --- # Step 1: I separated the research into information types The original report was not treated as one giant block of text. It was separated into content families. ### Company information - History. - funding. - acquisitions. - leadership. - ownership. - revenue. - employees. - customers. - stock. - strategic headlines. ### Platform information - Network. - security. - Zero Trust. - compute. - databases. - storage. - AI. - agents. - media. - analytics. - privacy. - administration. ### Practical application information - Outsource Access internal hub. - client portal. - coach lineup tool. - Family Hub. - Brad’s Command Center. - healthcare-oriented system. ### Decision information - Competitors. - prices. - security responsibilities. - what Cloudflare can replace. - what it cannot replace. - recommended implementation sequence. Once those were separated, each type could receive the most useful presentation format. --- # Step 2: I turned prose into structured data Instead of hard-coding 128 product descriptions into 128 separate visual components, I converted them into structured records. Conceptually, each capability became something like: > **Name:** D1 > **Category:** Database and storage > **Plain-English explanation:** A managed serverless SQL database based on SQLite > **User-experience example:** A manager assigns an employee to a client, and every authorized dashboard immediately reflects that assignment The visual interface then reads those records and automatically creates: - Capability cards. - category counts. - search results. - filters. - detail windows. - global-search results. That is why the interface can feel like a full software product even though the content is packaged in a static website. The same method was used for customers, competitors, projects, funding rounds, leaders, acquisitions, historical events, and pricing presets. --- # Step 3: I matched each information type to an interaction This is the main design reasoning behind the guide. ## History is chronological So it became a timeline. ## Revenue and employees are numerical trends So they became charts. ## Cloudflare capabilities are a large catalog So they became searchable and filterable cards. ## Application ideas involve selecting requirements So they became an architecture builder. ## An AI agent consists of connected layers So it became a layered interactive diagram. ## Customer evidence involves browsing industries and products So it became a searchable customer directory. ## Pricing depends on quantities So it became a calculator with sliders and presets. ## Competitors vary depending on the problem So they became selectable comparison lenses. ## Security is not merely something to read So it became an actionable checklist stored in the browser. ## Outsource Access needs an implementation order So the recommendations became a staged roadmap. The interface was designed around **what the reader needs to do with the information**, not merely how the original report happened to be formatted. --- # Step 4: I created several levels of explanation The guide intentionally supports different levels of attention. ## Level 1: instant understanding The opening screen tells you: - What Cloudflare is. - what it should do for Outsource Access. - what should remain in Google. - where specialist providers remain useful. ## Level 2: scanning Headlines, cards, charts, labels, and architecture diagrams allow quick browsing. ## Level 3: exploration Search, filters, tabs, and dialogs expose more detail. ## Level 4: action The architecture builder, pricing estimator, roadmap, and checklist convert learning into planning. The original report mostly operates at Levels 2 and 3. The web guide adds Levels 1 and 4. --- # Step 5: I rewrote for mobile reading A paragraph that works in a research report often performs poorly on a 390-pixel-wide phone. So long paragraphs were converted into: - Short summaries. - cards. - expandable details. - concise labels. - single-purpose panels. - progressive disclosure. - swipeable or scrollable structures. - mobile navigation. - responsive stacked layouts. “Progressive disclosure” simply means that the interface shows the most important information first and lets you open additional detail when you need it. That prevents a mobile user from encountering an unbroken wall of 20,000 words. --- # Step 6: I built a real static web application The interface consists of four principal layers. ## `data.js` — the knowledge layer This contains structured content such as: - Capabilities. - history. - customers. - leaders. - funding. - projects. - competitors. - prices. - sources. ## `index.html` — the page structure This defines the major sections: - Overview. - history. - capabilities. - builder. - AI stack. - customers. - pricing. - competitors. - security. - roadmap. - limitations. - glossary. - sources. ## `styles.css` — the visual system This controls: - Typography. - spacing. - responsive breakpoints. - mobile layouts. - cards. - charts. - dialogs. - dark mode. - print formatting. - visual hierarchy. ## `app.js` — the behavior This controls: - Searching. - filtering. - modal windows. - pricing calculations. - architecture selection. - checklist persistence. - navigation. - dark mode. - mobile menus. - timeline controls. - interactive customer and competitor views. There is also: - A service worker for offline behavior. - a web-app manifest. - deployment configuration. - a self-contained HTML edition. - a Cloudflare-ready ZIP. --- # It looks like a sophisticated hosted platform, but it is technically very lightweight This is an important distinction. The current guide is a **static client-side application**. It does not require: - A database. - user accounts. - a backend server. - an AI model running every time you use it. - a paid API. - a Cloudflare Worker processing each interaction. The search, filters, calculator, checklist, and builder run directly inside your browser using JavaScript. That means it can be: - Hosted extremely cheaply. - opened offline. - copied easily. - deployed to Cloudflare as static assets. - loaded quickly. - used without transmitting private interaction data to a backend. It is interactive because the browser is manipulating structured content—not because a live AI agent is generating every page. That is also a beautiful example of why Cloudflare can be inexpensive: a sophisticated-looking application does not necessarily require an expensive server architecture. --- # A concrete example of the translation Consider the original explanation of **R2**. The written report describes: - What object storage is. - what R2 stores. - how private buckets work. - temporary signed access. - PDF security. - encryption. - possible Outsource Access uses. - pricing. - how it compares with Google Drive. - HIPAA considerations. In the guide, that same body of knowledge appears in multiple locations: 1. **R2 capability card** Explains R2 in plain English. 2. **Capability detail window** Gives a concrete use example. 3. **Architecture builder** Adds R2 when you select private files or PDFs. 4. **Internal operations blueprint** Uses R2 for controlled client and employee documents. 5. **Family Hub blueprint** Uses R2 for private household records. 6. **Security section** Explains that buckets should remain private and access should be temporary. 7. **Pricing estimator** Lets you model R2 storage volume. 8. **Security checklist** Asks whether private buckets, file authorization, classification, retention, and backups have been implemented. So one original topic was not just shortened into one card. It was **distributed across the interface wherever it becomes relevant**. That is a major reason the guide feels much more sophisticated than a formatted article. --- # Another example: your product ideas Your original request mentioned: - Replacing spreadsheets. - client dashboards. - a coach lineup tool. - a Family Hub. - Brad’s Command Center. - AI agents. - private PDFs. - HIPAA-sensitive applications. Instead of presenting those as six isolated essays, I extracted their common building blocks: - Identity. - tenant separation. - structured records. - private files. - workflows. - AI knowledge. - live updates. - browser automation. - audit logs. - external customer authentication. - healthcare controls. The architecture builder recombines those blocks according to the selected product. That is not copied prose. It is a **small decision engine derived from the research**. --- # The best description of the fidelity I would characterize it like this: | Dimension | Fidelity to original report | |---|---| | Central conclusions | Very high | | 128-product catalog | Very high | | Major company facts and numbers | High | | Customer and competitor coverage | High, but summarized | | Application recommendations | Very high | | Exact wording | Low by design | | Every paragraph and caveat | Partial | | Every pricing dimension | Partial | | Sentence-level citations | Lower than the report | | Interactivity and application value | Much higher than the report | So it is neither a transcript nor a different project. It is best described as: > **The same research corpus, re-authored as an interactive Cloudflare learning and strategy system.** --- # What a truly full-fidelity edition would look like A full-fidelity version could preserve the current experience while adding: - A **“Read the complete research”** mode. - Every original paragraph. - sentence-level or paragraph-level citations. - expandable “Deep Dive” panels behind every capability. - complete pricing tables beneath the calculator. - full customer case-study notes. - detailed stock, ownership, and financial analysis. - complete reliability and controversy history. - an indicator showing whether a statement is documented, inferred, or recommended. - a change log for future Cloudflare updates. That would create three products in one: 1. **Executive briefing** 2. **Interactive learning environment** 3. **Complete research archive** The current interface is primarily the first two. The original written response remains the more exhaustive narrative research layer. You can inspect the underlying structure in the [complete editable source project](sandbox:/mnt/data/Cloudflare-Expert-Guide-Source.zip), while the [interactive guide](sandbox:/mnt/data/cloudflare-expert-guide-published/cloudflare-expert-guide.html) shows how those structured records become the finished experience. --- # Turn 7 · User **Category:** Full-fidelity export request **Title:** Request for a complete categorized Cloudflare project containing the entire chat Okay I'm a little confused. What do I do here? Is the link above to the complete editable source project another zip file I can have deployed on Cloudflare that will show the comprehensive detail of all the original analysis, including what you just sent me, which explains your thinking? I want everything in this entire chat exportable to a Cloudflare project and categorized --- # Turn 8 · Assistant **Category:** Master export delivery **Title:** Delivery and instructions for the complete Cloudflare Master Knowledge Base # Yes—this is now the complete package you were asking for The earlier **Cloudflare Expert Guide Source ZIP** was only the editable source for the condensed interactive dashboard. It was not a verbatim archive of every paragraph in the original research or every message in the conversation. I have now created a new **Cloudflare Master Knowledge Base** that combines all of those layers in one project: - The original interactive Cloudflare Expert Guide - The complete original 16,700-plus-word research response - All 41 numbered research chapters, plus the opening scope and final assessment - The complete substantive user-and-assistant conversation through this master-export request - The full explanation of how the research was translated into the interactive interface - A categorized source-link library - Global search across research sections, capabilities, conversation turns, design rationale, and sources - Raw Markdown and JSON exports - Cloudflare deployment files, build scripts, and publishing instructions ## What to do 1. Download **Cloudflare-Master-Knowledge-Base-Site.zip** when you simply want to publish the finished site. 2. In Cloudflare, open **Workers & Pages**, create a Direct Upload project, and upload that ZIP or its unzipped folder. 3. Download **Cloudflare-Master-Knowledge-Base-Source.zip** when you want the editable project, content files, generator, and Wrangler configuration. 4. Keep the old Cloudflare Expert Guide ZIP only as the original first-edition artifact; it is no longer the complete package. The master home page makes the distinction explicit. It gives you separate entrances for the interactive guide, complete research, complete chat archive, design rationale, sources, and raw exports. The original research is preserved as delivered and labeled current through July 22, 2026. Cloudflare changes quickly, so time-sensitive prices, leadership, product availability, beta status, stock information, and compliance scope should be revalidated before important decisions.