OpenAI’s updated Agents SDK adds a model-native harness and native sandbox execution so agents can inspect files, run commands, edit code, and continue across longer tasks. It launches generally available in Python with support for sandbox providers including Blaxel, Cloudflare, Daytona, E2B, Modal, Runloop, and Vercel.
#sandbox
RSS FeedA Launch HN post with around 260 points introduced Freestyle as infrastructure for coding agents, highlighting sub-second VM startup, live forking of running sandboxes, pause-and-resume persistence, built-in git hosting, and full Linux VMs intended for agent platforms rather than lightweight demo containers.
Perplexity introduced Computer for Enterprise on March 12, 2026 as a managed execution layer for enterprise agent workflows. The company says it inherits SOC 2 Type II, SAML SSO, audit logs, and admin controls, while keeping browser activity and code execution inside isolated sandbox environments.
Vercel used X on March 12, 2026 to show how Notion Workers runs agent-capable code on Vercel Sandbox. Vercel's write-up says Workers handle third-party syncs, automations, and AI agent tool calls, while Sandbox provides isolation, credential management, network controls, snapshots, and active-CPU billing.
Perplexity said on March 11, 2026 that its Sandbox API will become both an Agent API tool and a standalone service. Existing docs already frame Agent API as a multi-provider interface with explicit tool configuration, so the update pushes code execution closer to a first-class orchestration primitive.