index.md

Agents Documentation

Getting Started

• Getting Started - Quick start guide for new users
• Adding to an Existing Project - Integrate agents into your app
• Understanding the Agent Class - Deep dive into the Agent class architecture

Core Concepts

• State Management - Managing agent state with setState(), initialState, and onStateChanged()
• Routing - How routeAgentRequest() and agent naming works
• HTTP & WebSockets - Request handling and real-time connections
• Callable Methods - The @callable decorator and client-server method calls
• Readonly Connections - Restricting which connections can modify state
• getCurrentAgent() - Accessing agent context across async calls

Client SDK

• Client SDK - Connecting from React (useAgent) and vanilla JS (AgentClient), state sync, and RPC calls

Communication Channels

• Email Service - Sending, receiving, and replying to emails
• Webhooks - Receiving and sending webhook events
• Push Notifications - Browser push notifications via Web Push API and scheduled delivery
• TODO: SMS - Text message integration (Twilio, etc.)
• Voice Agents - Build voice agents with real-time speech-to-text, text-to-speech, and conversation persistence
• TODO: Messengers - Slack, Discord, Telegram, and other chat platforms

Background Processing

• Queue - Immediate background task execution
• Scheduling - Delayed, scheduled, and cron-based tasks
• Retries - Automatic retries with exponential backoff and jitter
• Durable Execution - runFiber(), stash(), and crash recovery for long tasks
• Workflows - Durable multi-step processing with Cloudflare Workflows
• Human in the Loop - Approval flows and manual intervention

AI Integration

• TODO: AI SDK Integration - Using Vercel AI SDK with agents
• TODO: TanStack Integration - Using TanStack AI with agents
• Chat Agents - AIChatAgent class and useAgentChat React hook
• Server-Driven Messages - Autonomous agent workflows: scheduled follow-ups, queue processing, webhooks, chained reasoning
• TODO: Using AI Models - OpenAI, Anthropic, Workers AI, and other providers
• TODO: RAG (Retrieval Augmented Generation) - Vector search with Vectorize
• Sessions (Experimental) - Persistent conversation storage with tree-structured messages, context blocks, compaction, and search
• Workspace (Experimental) - Durable virtual filesystem backed by SQLite + R2
• Codemode (Experimental) - LLM-generated executable code for tool orchestration
• Client Tools Continuation - Handling tool calls across client/server
• Resumable Streaming - Automatic stream resumption on disconnect

Think (Experimental)

• Overview - Opinionated chat agent with built-in memory, tools, and streaming
• Getting Started - Build your first Think agent step by step
• Lifecycle Hooks - beforeTurn, onStepFinish, onChunk, onChatResponse, and more
• Tools - Workspace tools, code execution, extensions
• Client Tools - Browser-side tools, approvals, and concurrency
• Sub-agents and Programmatic Turns - RPC streaming, saveMessages, recovery

MCP (Model Context Protocol)

• Creating MCP Servers - Build MCP servers with McpAgent
• Securing MCP Servers - OAuth and authentication for MCP
• Connecting to MCP Servers - addMcpServer() and consuming external MCP tools
• MCP Transports - Transport options: Streamable HTTP, SSE, and RPC

Authentication & Security

• TODO: Securing your Agents - Authentication, authorization, and access control
• Cross-Domain Authentication - Auth across different domains

Observability & Debugging

• Observability - Monitoring and tracing agent activity
• TODO: Testing - Unit tests, integration tests, mocking agents
• TODO: Evals - Evaluating AI agent quality and behavior

Agent Studio

• TODO: Agent Studio - Local dev tool for inspecting and interacting with agent instances

Compute Environments

• Browse the Web (Experimental) - Full CDP access for web inspection, scraping, and debugging
• TODO: Cloudflare Sandboxes - Isolated environments for coding agents, ffmpeg, and heavy compute

Advanced Topics

• Long-Running Agents - Building agents that persist for weeks or months: lifecycle, recovery, async operations, and planning
• TODO: SQL API - Using this.sql for direct database queries
• TODO: Memory & Persistence - Long-term storage patterns
• Configuration - wrangler.jsonc setup, types, secrets, and deployment

Migration Guides

• Migration to AI SDK v5
• Migration to AI SDK v6

Reference

• TODO: API Reference - Complete API documentation
• TODO: FAQ / How is this different from Durable Objects?
• TODO: Resources & Further Reading

---

Contributing

Found something missing? Documentation contributions are welcome!