Every atomic assertion extracted from the underlying record, ranked by evidence strength.
Cloudflare and Anthropic have collaborated to integrate Claude Managed Agents with Cloudflare Sandboxes.
Anthropic describes the integration as "decoupling the brain from the hands."
The core agent loop runs in Anthropic (the "brain").
The infrastructure for running and executing code (the "hands") can be run anywhere, including Cloudflare.
Cloudflare is providing a faster and cheaper sandbox for Claude Agents.
The faster and cheaper sandbox executes code within a V8 isolate instead of a microVM.
An example custom tool could host a public file on Cloudflare's R2 object storage.
Users are encouraged to fork the repo to easily add customized tools.
Other built-in tools include call_service, which uses Cloudflare Mesh or Workers VPC to connect to private services.
Possibilities include hosting applications written on the fly with Dynamic Workers.
Users don't have to worry about infrastructure or scaling.
Users just write a few lines of code and hit deploy.
Cloudflare and Anthropic are working together to bring Cloudflare's flexibility, scale, and security to more users.
The new integration makes it easy to run tens of millions of agents using isolates.
Possibilities include running edge inference with Workers AI.
image_generate pairs well with Claude providing text-based inference.
Customizable proxies allow better observation of how agents interact with the outside world.
The "Getting Started with Managed Agents guide" helps set up Claude Managed Agents with Cloudflare in minutes.
The new integration gives users more control over agent sandboxes.
The new integration secures connections to private services.
The new integration improves observability for agent sandboxes.
Cloudflare's Developer Platform has expanded in the past year to give more developers tools to run agents at scale.
Cloudflare's Developer Platform includes Sandboxes for full stateful Linux microVMs at scale.
Cloudflare's Developer Platform includes an Agents SDK, providing a simple and customizable agent framework.
Cloudflare's Developer Platform includes Browser Run, which gives agents fully programmable and observable browsers.
Cloudflare's Developer Platform includes Dynamic Workers, allowing for dynamic sandboxed code execution at massive scale.
Cloudflare's goal is to make Cloudflare the simplest, most secure, and most programmable cloud for agents.
Integrating with Claude Managed Agents is another step in Cloudflare's direction to become the simplest, most secure, and most programmable cloud for agents.
Users can run their agent loop on the Claude Platform.
Users can use Cloudflare to execute code for Claude Managed Agents.
Possibilities include giving each agent session a git-backed repo with Artifacts.
The default deployment template provides enhanced security by running all agent traffic through customizable proxies.
A default deployment template has been created to get started in minutes.
The new integration makes it easy to write custom tools that take advantage of Cloudflare.
Customizable proxies allow secure injection of credentials.
Customizable proxies prevent data exfiltration.
Users can use Cloudflare to run custom tool calls for Claude Managed Agents.
The default deployment template provides sandbox control and observability.
Users can get detailed sandbox metrics and logs.
Users can SSH into running machines.
Users can customize sandbox images.
The default deployment template provides lightweight sandboxes.
Lightweight sandboxes allow writing and executing untrusted code in a traditional microVM or a lightweight isolate.
Lightweight isolates enable massive scale.
Lightweight isolates boot sandboxes in milliseconds.
Lightweight isolates minimize infrastructure spend.
The default deployment template provides private service connectivity.
Agents can connect to private internal services without exposing them to the Internet.
The default deployment template provides Browser Control and Observability.
Users can get an audit trail of every agent's browser sessions.
Browser Control and Observability includes session recording.
Browser Control and Observability includes human-in-the-loop flows.
The default deployment template provides email functionality for agents.
Each agent can be given its own email address.
The default deployment template provides custom tools.
Users can extend agents with tools without needing additional infrastructure.
Users just write functions and deploy to create custom tools.
Claude Managed Agents allow developers to easily define and run agents on the Anthropic platform.
In Anthropic's managed environments, Claude can read files, run commands, browse the web, and execute code.
Users can use Cloudflare to secure connections for Claude Managed Agents.