Cloudflare, Inc., the leading connectivity cloud company, announced several new offerings designed to accelerate AI agent development. With the industry’s first remote Model Context Protocol (MCP) server, Cloudflare enables developers to easily build and deploy high-performance AI agents, provides widely accessible Durable Workflows, and offers a free-tier Durable Objects offering. These offerings enable developers to build agents in minutes instead of months, easily, affordably, and at scale.
AI agents—AI-powered systems that can act autonomously, make decisions, and adapt to new environments—represent the future of AI. AI agents hold the potential to unlock massive productivity gains, yet companies are struggling to develop agents that deliver real returns. Developing such agents requires access to three core components: AI reasoning models, workflows for execution, and APIs for accessing tools and services. To develop scalable agentic systems, companies need access to a platform that can deliver all three components in a scalable and cost-effective manner.
“Cloudflare is the best environment for developing and scaling AI agents. Period. The most innovative companies out there understand that agents are the next big step in applying AI, and they choose Cloudflare because we have everything they need to build quickly and at scale on our Workers platform,” said Matthew Prince, co-founder and CEO of Cloudflare. “Cloudflare has zeroed in on this moment: First, we built the most interconnected network on the planet. Then we built a developer platform that leverages that network to run code from 95% of the people online within 50 milliseconds. And we’re continuing to accelerate to give developers the best tools to build agentic AI.”
Also Read: phoenixNAP Advances Cloud Services Using HPE Disaggregated Data Center Modular Hardware System Servers with Intel Xeon 6
With today’s announcement, Cloudflare’s developer platform addresses some of the biggest challenges in developing AI agents with the following innovations:
Enabling intelligent, autonomous actions with the industry’s first remote MCP server
MCP is a rapidly growing open-source standard that allows AI agents to interact directly with external services. This allows AI to move from issuing simple instructions to actually performing tasks at the user’s request—such as sending an email, booking a meeting, or implementing code changes. Until now, MCP was limited to running locally on a device, making the standard accessible to developers and early adopters but preventing wider mainstream adoption.
Cloudflare will make it easy to develop and deploy remote MCP servers on Cloudflare, allowing any AI agent to securely connect over the internet and interact with services like email without the need for a locally hosted server. MCP servers built on Cloudflare can store context, providing each user with a persistent, ongoing experience. Through partnerships with Auth0 , Stytch , and WorkOS , Cloudflare also simplifies authentication and authorization, allowing users to delegate permissions to agents and dramatically simplifying the deployment of secure agents.
Developing intelligent, context-aware AI agents with Durable Objects now on a free-tier basis
Durable Objects , previously available only through paid subscriptions, are now available to developers as a free-tier offering from Cloudflare, expanding broad and democratized access to a critical component for agent development. Durable Objects are a special type of Cloudflare worker that combines compute and storage capabilities, enabling stateful applications to be developed in a serverless operating environment without infrastructure management. Durable Objects are the ideal foundation for AI agents that need to maintain context across interactions—for example, remembering past preferences or adapting their behavior based on previous events. Cloudflare’s network ensures that Durable Objects can scale to millions of simultaneous user interactions and serve agents close to the original request, ensuring each customer receives a fast, low-latency response.
Deploy persistent, multi-step applications with workflows, now generally available
Workflows allow you to build multi-step applications that can automatically retry, persist, and run for minutes, hours, days, or weeks. Workflows is now generally available and provides developers and businesses with a reliable way to build and manage multi-step applications powered by AI. For example, building an agent workflow for booking travel would require searching for flights within a specific price range, which would require persistent searching over a specified period of time. Once the flights are found, an agent would purchase the tickets using the traveler’s information and a credit card. The confirmation would then be sent to all travelers in the group.
Pay only for what is used to achieve the most cost-effective AI deployment
AI inference, unlike training, is difficult to predict and inherently inconsistent. It relies on human behavior and depends, among other things, on the time of day and the action a person intends to take. With traditional hyperscalers, companies must therefore prepare and plan to provide the highest capacity they can expect, even if that capacity is only reached during peak hours. Cloudflare’s serverless platform, in contrast, automatically scales inference and AI agent resources as needed in milliseconds, from zero to global scale. This ensures that user organizations only pay for what they use, drastically reducing costs compared to traditional cloud implementations that require constant provisioning.
“Cloudflare offers a developer-friendly ecosystem for building AI agents, including a free-tier offering for Durable Objects and serverless options for AI inference,” explains Kate Holterhoff, Senior Analyst at RedMonk. “These low-cost and easy-to-use options could enable more companies to adopt and experiment with agentic AI.”
Cloudflare is a leader in making AI inference accessible and removing barriers that have kept AI out of reach for most enterprises. Cloudflare operates GPUs in more than 190 cities worldwide, bringing low-latency AI as close to the user as possible.
Source: BusinessWire