Deploys stateful, real-time AI agents using the Cloudflare Agents SDK on Workers and Durable Objects.
This skill provides specialized guidance for building autonomous, stateful AI agents on Cloudflare's serverless infrastructure. It streamlines the decision-making process between using the lightweight Vercel AI SDK for simple chat flows and the robust Cloudflare Agents SDK for complex applications requiring bidirectional WebSockets, persistent SQLite state, and multi-agent orchestration. By automating infrastructure setup, Durable Object bindings, and SQLite migrations, this skill enables developers to build scalable, long-running agentic workflows with built-in human-in-the-loop patterns and Model Context Protocol (MCP) support.
Key Features
01Architectural guidance for selecting between AI SDK and Agents SDK
02Real-time bidirectional communication setup using WebSockets
03Persistent state management with built-in SQLite (up to 1GB per agent)
04Autonomous task scheduling for cron-based and delayed workflows
0591 GitHub stars
06Standardized patterns for MCP server implementation and tool execution
Use Cases
01Developing real-time collaborative AI agents with persistent memory
02Orchestrating multi-agent systems for complex asynchronous workflows
03Building stateful MCP servers for enhanced tool-use capabilities
