#sandboxing

All articles tagged with "sandboxing"

Building a safe, effective sandbox to enable Codex on Windows

OpenAI details the Windows sandbox built to constrain Codex

OpenAI says it built a custom Windows sandbox for Codex after finding that native isolation tools could not enforce the same guarded file and network boundaries available on macOS and Linux.

Key Takeaways

OpenAI says Codex on Windows previously forced users into frequent approvals or unrestricted access.
The company built a custom sandbox because existing Windows isolation tools did not meet its requirements.

DT Editorial Team·May 13, 2026·via openai.com

OpenAI Adds Sandboxed Execution to Its Agents SDK

OpenAI has updated its Agents SDK with native sandbox support, new tool integrations, and a manifest system aimed at making agent workflows safer and more durable.

Key Takeaways

OpenAI added native sandbox support to its Agents SDK.
The update includes MCP tool usage, shell execution, and apply-patch editing.

DT Editorial Team·Apr 16, 2026·via the-decoder.com

OpenAI Pushes Agents Closer to Production With Native Sandboxes and a Model-Native Harness

OpenAI has expanded its Agents SDK with native sandbox execution and a model-native harness, a move aimed at making long-running, tool-using agents easier to build and safer to operate.

Key Takeaways

OpenAI added native sandbox execution and a model-native harness to the Agents SDK.
The update targets long-running agent workflows involving files, tools, and code edits.

DT Editorial Team·Apr 16, 2026·via openai.com