Cloudflare Launches Sandboxes to Give AI Agents a Persistent Computer

Cloudflare

Apr 13, 2026 · Updated Apr 25, 2026

Cloudflare Sandboxes are now generally available, providing AI agents with isolated, persistent environments to execute code and manage development servers. This infrastructure shift allows agents to perform multi-step engineering tasks like human developers while maintaining security through programmable credential injection.

Cloudflare, a network and security company providing developer platform services, moved its Sandboxes and Containers to general availability. The update introduces PTY support (a real terminal interface) and persistent code interpreters (stateful execution environments) to give AI agents a persistent environment. These environments start on demand and sleep when idle.

Most agent systems rely on request-response loops that lose state between turns. By providing a full computer with a filesystem and background processes, this infrastructure enables agents to participate in tight feedback loops—like running tests and fixing errors. Secure credential injection allows agents to access private services without seeing tokens.

You can now deploy up to 15,000 concurrent instances using a pricing model that only charges for active CPU cycles. Snapshot support will allow you to save full disk states to R2 for instant session restoration. The SDK is available via npm for immediate integration.

View the full update on blog.cloudflare.com

Cloudflare

@CloudflareApr 13

AI agents need more than a prompt—they need a computer. 💻 Cloudflare Sandboxes are now GA. Give your agents a secure, persistent environment to clone repos, run Python/JS, and debug via a real PTY terminal. #AgentsWeek https://t.co/S1CzYIfpNl

1367

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

See all AI news & updates from Cloudflare →

Keep reading

Cloudflare Launches Outbound Workers to Secure AI Agent Network Access

Cloudflare introduced Outbound Workers for Sandboxes, a programmable proxy that lets developers inject credentials and enforce security policies at the network level. This allows AI agents to interact with sensitive services without ever having direct access to the raw authentication tokens.

Vercel Sandbox Adds Persistent Drives for AI Agents to Retain Memory

VercelJun 5

Vercel Sandbox Adds Persistent Drives for AI Agents to Retain Memory

Vercel Sandbox now supports persistent drives in private beta. This separates storage from compute, allowing AI agents to retain their workspaces and "memories" across disposable sandbox instances. This is crucial for building reliable, multi-step agentic systems that need to maintain state.

OpenAI Secures Agent Workflows by Separating Credentials from Execution Environments

OpenAIApr 15

OpenAI Secures Agent Workflows by Separating Credentials from Execution Environments

OpenAI updated its Agents SDK to support isolated execution environments through a native Vercel Sandbox integration. This architecture keeps sensitive credentials in the orchestration layer while allowing agents to run code in secure MicroVMs. It enables developers to build autonomous systems that can safely execute untrusted code without risking host system exposure.

Guillermo RauchMay 20

Vercel and Anthropic Secure Autonomous Agents via Firewall Level Credential Injection

Anthropic launched self-hosted sandboxes for Claude Managed Agents alongside a dedicated Vercel Sandbox integration. This architecture keeps sensitive API keys outside the execution environment, allowing autonomous agents to safely interact with private infrastructure without the risk of credential leakage.