Cloudflare unveils major AI Agent development tools

Today, Cloudflare unveiled a comprehensive suite of developer tools focused on building autonomous AI agents as part of its Developer Week 2025 event that began on April 6. The announcements include the acquisition of database company Outerbase, the general availability of Cloudflare Workflows, the launch of AutoRAG, and significant enhancements to its Model Context Protocol capabilities, collectively strengthening Cloudflare's position as a platform for agentive AI development.

The timing of these announcements marks a significant moment in the developer landscape. Coming just one day after the start of Cloudflare's Developer Week 2025, these releases represent the company's strategic focus on enhancing its AI development ecosystem.

Cloudflare's acquisition of Outerbase represents a key component of this strategy. According to Brandon Strittmatter, co-founder of Outerbase, "This is such an amazing opportunity for us, and I want to explain how we got here, what we've built so far, and why we are so excited about becoming part of the Cloudflare team."

The acquisition aims to integrate Outerbase's database management capabilities with Cloudflare's existing tools, including D1 and Durable Objects. Outerbase has built technologies that simplify database access, visualization, and management - critical components for developers building data-driven AI applications.

"Databases are key to building almost any production application: you need to persist state for your users (or agents), be able to query it from a number of different clients, and you want it to be fast," Strittmatter explained.

Particularly noteworthy is Outerbase's Starbase, a SQLite-compatible database built on top of Cloudflare's Durable Objects, which provides automatic REST APIs, row-level security, and WebSocket support for streaming queries.

According to Matt Silverlock, one of the announcement authors, "We've already done a lot of thinking about how we're going to bring the best parts of Outerbase into D1, Durable Objects, Workflows, and Agents, and we're going to a share a little about what will be landing over the course of Q2 2025 as the Outerbase team gets to work."

The Outerbase cloud service will shut down on October 15, 2025, giving customers six months to transition, though the company will maintain open-source repositories for self-hosting options.

Simultaneously, Cloudflare announced the general availability of its Workflows product, a durable execution engine built on Cloudflare Workers. After a period in beta, Workflows now offers production-ready capabilities for building resilient, multi-step applications.

"Betas are useful for feedback and iteration, but at the end of the day, not everyone is willing to be a guinea pig or can tolerate the occasional sharp edge that comes along with beta software," noted Sid Chatterjee and Matt Silverlock in their announcement.

Workflows implements a step-based architecture where each step in an application is independently retriable, with state automatically persisted between steps. This makes it particularly useful for complex business processes like e-commerce order processing that might involve checking inventory, processing payments, and sending confirmations.

A significant enhancement to Workflows is the new waitForEvent API, which allows applications to pause execution until receiving an external trigger, such as human approval or a webhook from another service. This capability facilitates human-in-the-loop processes critical for many AI agent applications.

Cloudflare also introduced AutoRAG, a fully managed Retrieval-Augmented Generation pipeline. According to Anni Wang, AutoRAG simplifies how developers integrate context-aware AI into their applications by automating the process of ingesting, transforming, and storing data as vectors for semantic search.

"Building a RAG pipeline is a patchwork of moving parts. You have to stitch together multiple tools and services — your data storage, a vector database, an embedding model, LLMs, and custom indexing, retrieval, and generation logic — all just to get started," Wang explained.

AutoRAG abstracts this complexity by providing an end-to-end pipeline that automatically handles the entire process, continuously monitoring data sources to keep AI systems current without manual intervention.

The system works through two main processes: indexing, which runs in the background to process content into vectors optimized for semantic search, and querying, which retrieves relevant content when users make requests. During the open beta period, AutoRAG is free to enable, with compute operations for indexing, retrieval, and augmentation incurring no additional cost.

For developers building AI agents, Cloudflare has enhanced its Model Context Protocol (MCP) capabilities. The company's Agents SDK now supports building remote MCP clients with built-in transport and authentication functions, allowing AI agents to connect to external services.

"AI agents can now connect to and interact with external services through MCP. We've updated the Agents SDK to allow you to build a remote MCP client into your AI agent, with all the components — authentication flows, tool discovery, and connection management — built-in for you," according to Rita Kozlov, Dina Kozlov, and Vy Ton.

These enhancements enable agents to prompt users for access to third-party services, use tools from these services on users' behalf, call MCP servers from scheduled tasks, and connect to multiple MCP servers to discover new capabilities.

Cloudflare has also introduced integrated authentication solutions for MCP servers, partnering with Stytch, Auth0, and WorkOS to simplify securing agent interactions. Additionally, the company added support for WebSockets Hibernation to the McpAgent class, allowing MCP servers to sleep when inactive and instantly wake when needed, resulting in more efficient resource usage.

In a move that lowers barriers to entry for developers, Cloudflare announced that Durable Objects—a stateful computing service previously available only on paid plans—is now accessible on the free tier. This is particularly significant for agent development, as Durable Objects provide the stateful coordination needed for AI agents.

"To help you build AI agents on Cloudflare, we're making Durable Objects available on the free tier, so you can start with zero commitment," the company stated.

The free tier includes substantial allocations for compute and storage: 100,000 Durable Object requests per day, 13,000 GB-seconds of compute duration daily, and 5 GB of SQL storage. These resources enable developers to build and test meaningful applications before committing to paid plans.

These announcements collectively represent Cloudflare's comprehensive response to the growing demand for tools that simplify AI agent development. As Rita Kozlov noted in her kickoff post for Developer Week 2025, "It's an exciting time to be a developer. In fact, as a developer, the past two years might have felt a bit like every week is Developer Week."

The timing appears deliberate, as the AI landscape continues to evolve at a rapid pace. Kozlov observed that recent months have seen significant developments, including DeepSeek challenging assumptions about model training, MCP introducing new standards for LLM interfaces, and OpenAI's o4 model making significant strides.

For marketing professionals, these developments have substantial implications. The emergence of accessible agent-building platforms enables marketers to develop more sophisticated, automated customer engagement systems. AutoRAG's ability to efficiently process and retrieve information from marketing content can enhance knowledge management and customer support applications. Meanwhile, the database visualization and management capabilities from Outerbase could improve analysis of marketing data and campaign performance metrics.

The Agents SDK with MCP client support potentially enables marketing teams to build AI agents that can interact with multiple marketing platforms and data sources on behalf of users, streamlining complex multi-channel workflows. The adoption of Workflows for durable execution could provide more reliable automation of marketing processes that involve multiple steps and external service interactions.

As these technologies mature, marketing professionals will likely see opportunities to deploy increasingly autonomous systems that can handle complex tasks like content optimization, audience targeting, and campaign management with less direct human intervention.

With these announcements, Cloudflare continues to position itself as a comprehensive platform for developers building applications in the AI era. The company's approach focuses on solving the infrastructure challenges that can slow development, allowing developers to concentrate on creating value through their applications rather than managing underlying systems.

Timeline

• April 6, 2025: Rita Kozlov kicks off Cloudflare's Developer Week 2025 with introductory blog post.
• April 7, 2025: Cloudflare announces acquisition of Outerbase to enhance database and agent developer experience.
• April 7, 2025: Cloudflare Workflows reaches general availability with new waitForEvent API for human-in-the-loop processes.
• April 7, 2025: AutoRAG enters open beta, offering fully managed Retrieval-Augmented Generation capabilities.
• April 7, 2025: Enhanced MCP capabilities announced, including authentication integrations with Stytch, Auth0, and WorkOS.
• April 7, 2025: Durable Objects becomes available on the free tier to lower barriers to agent development.
• October 15, 2025: Scheduled shutdown date for the hosted Outerbase cloud service.