Help output is identical between 2.1.56 and 2.1.59 — no new flags or commands introduced.

Notable existing flags (already in 2.1.56):

• --tmux — Create a tmux session for the worktree (requires --worktree)
• --tools — Specify available tools from the built-in set
• --verbose — Override verbose mode setting from config

No migration required. Version bump only.

CLI Changes (discovered via --help diff)

Removed flag (breaking change):

• --disable-parallel-tools-execution — flag has been removed from the CLI

Config setting removed:

• parallel_tool_execution — configuration setting removed from config help topic

This suggests parallel tool execution is now always enabled and no longer configurable.

config help diff (0.0.417 → 0.0.418)

- `parallel_tool_execution`: whether to enable parallel execution of tools; defaults to `true`.
-   - Can be disabled with --disable-parallel-tools-execution flag

environment help

No changes detected between 0.0.417 and 0.0.418.

If any workflow configuration uses --disable-parallel-tools-execution, this flag must be removed as it no longer exists in 0.0.418. Parallel tool execution is now always active.

Release Highlights (from GitHub rust-v0.105.0)

New Features

• TUI syntax-highlights fenced code blocks and diffs, adds /theme picker with live preview, and uses better theme-aware diff colors (feat(tui): syntax highlighting via syntect with theme picker openai/codex#11447, feat(tui): add theme-aware diff backgrounds with capability-graded palettes openai/codex#12581)
• Voice dictation via spacebar hold to record/transcribe (experimental; set features.voice_transcription = true) (voice transcription openai/codex#3381)
• Multi-agent: spawn_agents_on_csv for CSV fan-out with built-in progress/ETA; sub-agents easier to follow with nicknames, cleaner picker, visible child-thread approval prompts (Agent jobs (spawn_agents_on_csv) + progress UI openai/codex#10935, feat: add nick name to sub-agents openai/codex#12320, feat: cleaner TUI for sub-agents openai/codex#12327, feat: better agent picker in TUI openai/codex#12332, feat: keep dead agents in the agent picker openai/codex#12570, Display pending child-thread approvals in TUI openai/codex#12767)
• New TUI commands: /copy (copies latest assistant reply), /clear and Ctrl-L (clear screen without losing thread; /clear also starts fresh chat) (feat(tui) /clear openai/codex#12444, tweaked /clear to support clear + new chat, also fix minor bug for macos terminal openai/codex#12520, feat(tui) - /copy openai/codex#12613, ctrl-L (clears terminal but does not start a new chat)  openai/codex#12628)
• Approval controls: Codex can request extra sandbox permissions per command; auto-reject specific approval prompt types without disabling all approvals (feat(core) Introduce Feature::RequestPermissions openai/codex#11871, feat: add Reject approval policy with granular prompt rejection controls openai/codex#12087)
• App-server: thread/list can search by title; thread status in read/list responses; thread/resume returns latest turn inline (app-server: improve thread resume rejoin flow openai/codex#11776, app-server: expose loaded thread status via read/list and notifications openai/codex#11786, feat: add search term to thread list openai/codex#12578)

Bug Fixes

• Long TUI links stay clickable when wrapped; fixes related clipping/layout issues (fix(tui): preserve URL clickability across all TUI views openai/codex#12067)
• Queued-message editing works in more terminals; follow-up prompts no longer stuck on Enter during streaming; approval dialogs respond with correct request id (fix(tui): queued-message edit shortcut unreachable in some terminals openai/codex#12240, fix(tui): queue steer Enter while final answer is still streaming to prevent dead state openai/codex#12569, fix: chatwidget was not honoring approval_id for an ExecApprovalRequestEvent openai/codex#12746)
• @ parsing more reliable in chat composer — commands like npx -y @scope/pkg@latest no longer trigger file picker (Fix @mention token parsing in chat composer openai/codex#12643)
• App-server websocket: thread listeners survive disconnects; Ctrl-C waits for in-flight turns; permessage-deflate clients can connect (app-server: retain thread listener across disconnects openai/codex#12373, codex-rs/app-server: graceful websocket restart on Ctrl-C openai/codex#12517, app-server: fix connecting via websockets with Sec-WebSocket-Extensions: permessage-deflate openai/codex#12629)
• Linux sandboxed commands get a minimal /dev, fixing entropy device failures (fix(linux-sandbox): mount /dev in bwrap sandbox openai/codex#12081)
• js_repl reports uncaught kernel failures clearly and recovers cleanly (fix(js_repl): surface uncaught kernel errors and reset cleanly openai/codex#12636, Fix js_repl view_image attachments in nested tool calls openai/codex#12725)

Documentation

• Added public security policy with Bugcrowd reporting guidance (docs: add codex security policy openai/codex#12193)

No migration required. All changes are additive features and bug fixes.

Voice transcription requires opt-in via features.voice_transcription = true in config.