Cloudflare Agents SDK

Description

The Cloudflare Agents SDK is designed for developers looking to build and host stateful AI agents. These agents can maintain persistent memory, utilize real-time WebSocket connections, and execute scheduled tasks. With the SDK, developers can create agents that interact through various communication channels, including chat, voice, email, and Slack, ensuring a versatile approach to user engagement.

When deploying agents on Cloudflare, each session is assigned a durable identity, allowing for local SQL storage and recoverable execution. This infrastructure eliminates the need for managing complex backend systems, as Cloudflare handles the scaling and deployment across its global network. Developers can focus on building their agents without worrying about session reconstruction or state externalization.

The SDK comprises four main components: communication channels, the agent harness, the SDK runtime, and tools. Communication channels define how users and systems interact with the agents. The agent harness manages the operational loop, including model calls, tool selection, and response handling. The SDK runtime provides essential infrastructure, including the Agent class, state management, routing, and observability features. Finally, tools enhance agent capabilities, allowing for browser automation, sandboxed code execution, and AI search functionalities.

Developers can create various types of agents, such as a streaming AI chat agent with human-in-the-loop approvals, a Slack command responder, a real-time voice agent, or an email routing agent. The SDK also supports integration with multiple AI providers, enabling flexibility in model selection. Overall, the Cloudflare Agents SDK offers a robust framework for building intelligent, responsive agents that can operate efficiently in diverse environments.

Cloudflare Agents SDK's Core Features

• Persistent memory

• Real-time WebSocket connections

• Scheduled tasks

• Durable identity

• Local SQL storage

• Global network deployment

• Multiple communication channels

• Agent harness for operational control

• Flexible tool integration

• No API keys required

Getting Started with Cloudflare Agents SDK

1. 1

   Install via package manager: Download and install the Cloudflare Agents SDK.

2. 2

   Configure: Set up your agent's communication channels and tools.

3. 3

   Build: Develop your agent using the SDK runtime and agent harness.

4. 4

   Deploy: Launch your agent on Cloudflare's global network.

5. 5

   Optimise: Monitor performance and make adjustments as necessary.

Cloudflare Agents SDK's Use Cases

• Streaming AI Chat Agent
• Slack Command Responder
• Real-time Voice Agent
• Email Routing Agent
• Browser Automation Agent

FAQ from Cloudflare Agents SDK

• What is the Cloudflare Agents SDK?

  The Cloudflare Agents SDK is a framework for building stateful AI agents with persistent memory and real-time connections.

• How do I deploy agents using the SDK?

  You can deploy agents by installing the SDK, configuring your agent, and launching it on Cloudflare's network.

• What communication channels are supported?

  The SDK supports chat, voice, email, Slack, and webhooks as communication channels.

• Is there a need for API keys?

  No API keys are required; the starter uses Workers AI by default.

• Can I integrate third-party AI models?

  Yes, you can swap in models from OpenAI, Anthropic, Google Gemini, or other providers.

• What tools can I use with the SDK?

  The SDK includes tools for browser automation, sandboxed code execution, AI search, and payments.

• What is the agent harness?

  The agent harness manages how the agent calls models, selects tools, and handles responses.

• How does the SDK handle scalability?

  Cloudflare runs your agents across its global network, scaling to millions of instances automatically.