Codex CLI 0.169 adds side chats and hardens untrusted execution
OpenAI's 13 November stable 0.168 and 0.169 train adds model-provider management, persistent multi-tab side chats, searchable older tasks, custom-gateway device authorization and an opt-in Cloud connection. Codex exec now keeps project configuration and MCP servers disabled in untrusted directories unless the operator explicitly saves trust.
Why it matters: The release makes parallel terminal work and provider onboarding easier while tightening a high-consequence trust boundary for non-interactive execution. Saved provider identity, explicit directory trust and side-chat persistence now need to survive restart…