New Features

Thread Forking: You can now fork a thread into sub-agents, making it easier to branch work without leaving the current conversation. This enables parallel workflows and experimentation within the same session.

Multimodal Custom Tools: Custom tools can now return multimodal output, including structured content like images, instead of being limited to plain text responses. This significantly expands the capabilities of tool integrations.

Realtime Voice Improvements: Realtime voice sessions now let you pick microphone and speaker devices, persist those choices, and send audio in a format better aligned with transcription—improving the user experience for voice-based interactions.

Model Availability Metadata: The app server now exposes richer model availability and upgrade metadata, with the TUI using it to explain plan-gated models with limited-run tooltips for better transparency.

Memory Configuration: Memories are now fully configurable, and a new codex debug clear-memories command lets you fully reset saved memory state when needed.

Bug Fixes

• Reconnecting with thread/resume now restores pending approval and input requests instead of leaving clients out of sync
• thread/start no longer blocks unrelated app-server requests, reducing stalls during slow startup paths such as MCP auth checks

Installation

Update to the latest version with:

npm install -g @openai/codex@0.107.0