Codex CLI 0.164.0 adds presets, enterprise MCP auth, and credential masking

OpenAI has released Codex CLI 0.164.0, an update that affects everyday terminal workflows, enterprise MCP connections, and remote-execution safety. It introduces ten numbered /presets slots for reusable prompts and slash commands, with a keyboard menu for quick access. /model and /permissions can now accept model, reasoning-effort, speed, and permission-profile settings inline. Terminals that support Kitty or Sixel graphics can also preview viewed and generated images in the fullscreen transcript.

For integrations, the official notes say enterprise MCP connections can authenticate with eligible Codex account grants or a dedicated sign-in, including OAuth clients that require secrets. For remote execution, supported remote Bash commands and Zsh snapshot sessions mask credentials, including credentials from custom providers. Those protections reduce the chance that sensitive values surface in logs and session records; they do not replace access controls for secrets, permission profiles, or the MCP server itself.

Version 0.164.0 also addresses reliability and isolation. A failed compaction no longer leaves partial summaries in conversation history, and resumed sessions locate compressed histories and restore saved thread settings. Repeated -c or --config overrides across subcommand levels are preserved. The release hardens sandbox checks for case-variant metadata paths on macOS and for replaceable symlinks used to locate executables.

Upgrade first in a non-production workspace. Teams using /presets should check that shared prompts still fit their approval policies; teams with enterprise MCP should test their OAuth sign-in flow; and remote-execution users should verify that redacted logs remain sufficient for debugging. Run a regression pass for resumed sessions and automation commands if you rely on compressed histories or layered CLI configuration.

Related catalog entry