Cloudflare Agents

Cloudflare Agents is a platform for building agentic AI workflows on Cloudflare. Connect input, model access, durable execution, and tools on one stack.

AI 에이전트 개발 AI 개발 도구 프로그래밍 에이전트

웹사이트 방문

Overview

Cloudflare Agents is a platform for building agentic AI on Cloudflare. The home page describes agents as systems that can take user input, reason about a task, keep state across steps, call an LLM, and then use tools to complete work.

The product is presented as a Cloudflare-native stack for agent workflows. The page ties together input channels such as email, chat, and voice with Cloudflare services for inference, execution, storage, and integrations, so teams can build the full flow on one platform.

The source also emphasizes Cloudflare’s runtime model for agents: pay for CPU time rather than wall time, support long-lived connections through WebSocket hibernation, and use Durable Objects and Workflows to keep an agent moving through multi-step jobs.

The collected pricing URL returned "Not Found," so the source set does not confirm any public pricing page, plan structure, or billing tier for this product.

Features

Durable execution

Cloudflare positions the platform around durable execution so agent workflows can continue across multiple steps instead of stopping at a single prompt-response exchange.

LLM connectivity

The source says agents can use Cloudflare-hosted inference through Workers AI or connect to external model providers through AI Gateway.

State and workflow coordination

The page describes state and compute together as the execution layer, using Durable Objects for state and Workflows for orchestration.

Tool and action support

Agents can take structured actions through APIs, MCP servers, and utilities such as Browser Rendering, Vectorize, and D1.

Usage-based runtime behavior

The page notes that Cloudflare Workers bills CPU time rather than wall time, and it highlights WebSocket hibernation in Durable Objects for long-lived connections.

Use Cases

Multi-channel agent intake
Build agents that start from email, chat, or voice input and then move through planning, reasoning, and task completion without leaving the Cloudflare platform.
LLM-backed task automation
Create agents that connect to Workers AI or an external model provider through AI Gateway when the workflow needs model inference as part of the loop.
Long-running workflows with state
Use Durable Objects and Workflows when an agent needs durable execution, state, and the ability to revisit a plan as new information arrives.
Tool-enabled task completion
Wire agents to APIs, MCP servers, and Cloudflare utilities like Browser Rendering, Vectorize, or D1 when the job needs external actions or data access.
Persistent connection handling
Build agent runtimes that keep connections alive efficiently, especially where WebSockets and other long-lived interactions are part of the design.

Pros and Cons

Pros

Combines input, model access, execution, state, and tools in one Cloudflare-oriented workflow.
Supports both Cloudflare-hosted inference and connections to external LLM providers.
Includes building blocks for long-running agent tasks, including Durable Objects, Workflows, and WebSocket hibernation.
Mentions a range of practical tool surfaces such as APIs, MCP servers, Browser Rendering, Vectorize, and D1.

Cons

The collected pricing page is missing, so the source does not confirm plan names, limits, or billing details.
The source is high level and does not provide documentation on setup steps, supported outputs, or platform constraints.

FAQ

What is Cloudflare Agents?

Cloudflare Agents is a platform for building agents on Cloudflare. The home page describes a workflow that starts with user input, connects to an LLM, guarantees execution with state and compute, and then takes action through tools.

What Cloudflare products does it reference for building an agent?

The source points to email, chat, and voice as possible input channels, and it shows Cloudflare components such as Email Workers, WebSockets, Workers AI, AI Gateway, Durable Objects, Workflows, MCP servers, Browser Rendering, Vectorize, and D1 as parts of the agent-building flow.

How much does Cloudflare Agents cost?

The public site text does not show a pricing model or plan details. The pricing page path returned "Not Found," so there is no supported pricing information to summarize from the collected sources.

Who is Cloudflare Agents for?

The page frames Cloudflare Agents as a platform for developers building agentic AI workflows, including agents that need durable execution, LLM access, and tool use across multiple steps.

What does the source not confirm about the product?

The page emphasizes that agents can plan, iterate, and act, but the collected sources do not provide product documentation on setup constraints, output formats, or limits beyond that high-level workflow.