Codex CLI hardens sandbox, expands MCP capacity

This is the kind of release that matters more than a flashy model bump: agentic coding only scales when execution boundaries stay trustworthy. The broader MCP support is valuable, but it also raises the importance of strict server permissions and isolation.

• –Hook persistence and sandbox boundaries are coupled; configuration that can alter execution mode is a supply-chain risk.
• –Scheduler fixes should make background and multi-step agent workflows more reliable.
• –Expanded MCP capacity improves tool composition, but increases the blast radius of poorly scoped servers.
• –Codex remains an open-source, terminal-native coding agent built for real repository work. [OpenAI](https://openai.com/codex/), [Product Hunt](https://www.producthunt.com/posts/954333)