← All organizations

Cloud infrastructure, security, and developer platforms

Cloudflare

Cloudflare provides cloud services for securing websites, delivering content, connecting employees, and running applications. Its global network combines CDN, DDoS protection, web application firewall, and bot management services with Zero Trust controls for workforce access. Developers use Cloudflare Workers, storage services, and built-in AI inference to build applications and agents without managing servers. These services run on the same network, bringing application delivery, security, and development into one platform for startups and enterprises.

Founded in 2009 by Matthew Prince, Michelle Zatlyn, and Lee Holloway, Cloudflare grew out of Project Honey Pot’s work tracking Internet abuse and publicly launched in 2010. Prince remains CEO, while Zatlyn is president. Its engineering contributions include workerd, the open-source JavaScript and WebAssembly runtime underlying Workers. It supports web-standard APIs and runs multiple services in separate isolates within one process, reducing communication overhead. Developers can also use it to self-host Workers applications or test them locally.

Cloudflare combines free entry tiers with paid website plans, workforce subscriptions, and usage-based developer services. It reported that more than 45% of Fortune 500 companies were paying customers as of June 2026. Its completed acquisition of Replicate in December 2025, for $57.4 million in total cash purchase consideration, added a platform that lets developers deploy and run AI models.

www.cloudflare.com

Start here

  1. Building Agents (the hard parts!)

    Start with Rita Kozlov's talk to understand how voice interfaces, MCP integrations, authentication, and persistent state fit into an agent application.

    Rita KozlovAI Engineer World's Fair 2025

  2. Code Mode: Let the Code do the Talking

    Sunil Pai's Workers discovery demonstration makes the search-and-execute approach concrete, then introduces its sandbox and observability considerations.

    Sunil PaiAI Engineer Europe 2026

  3. Why, and how you need to sandbox AI-Generated Code? — Harshil Agrawal, Cloudflare

    Harshil Agrawal's talk connects hallucinations, credential exposure, and indirect prompt injection to decisions about execution capabilities.

    Harshil AgrawalAI Engineer Europe 2026

  4. MCP = Mega Context Problem - Matt Carey

    Cloudflare's Matt Carey explains why exposing every API endpoint as an MCP tool overwhelms agent context windows: he describes an OpenAPI specification of roughly 2.3 million tokens, approximately 2,600 endpoints, and fragmented product-specific MCP servers.

    Matt CareyAI Engineer Europe 2026

Messages from the stage

State and coordination for agents

Rita Kozlov connects clients, reasoning, workflows, and tools through CRM and enterprise-approval examples. Sunil Pai and Matt Carey discuss Durable Objects for persistent collaborative state, alongside agent filesystems backed by SQLite and R2.

API access within a context budget

Matt Carey examines why exposing every endpoint as an MCP tool overwhelms context windows. Sunil Pai presents typed API discovery and executable JavaScript as a way to compose operations without repeatedly exchanging JSON tool calls.

Isolation matched to the workload

Harshil Agrawal compares restricted V8 isolates with Linux-container sandboxes for workloads needing files, dependencies, and development servers. Kenton Varda's Gadgets demonstration extends isolation to generated client and server code using browser restrictions and controlled communication.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-27