Codex CLI v0.171.0 is a OpenAI release published 2026-11-20 with 353 listed changes, flagged for security fixes. Summary and full notes below, extracted from the official notes.
One free email: every breaking change, retirement and price change across 23 coding agents. No spam, unsubscribe anytime.
Subscribe free← All Codex CLI releases and breaking changes
## New Features - Added GPT-6.2 Sol and Astra to model catalogs, with Sol as the default catalog selection. (#62106) - Enabled the TUI workspace by default, with searchable document previews, image previews, and subagent transcripts alongside the main conversation. (#61352, #61888, #61889, #61919, #62157) - Added configurable undo and redo shortcuts to the non-Vim prompt composer, including restoration of attachments and large pastes. (#61954) - Added `codex exec --remote <ADDR>` to start or resume noninteractive sessions on an existing app server. (#61707) - Added `/rc` and `/remote` to manage local remote control and display device pairing codes from the TUI. (#62121) - Added custom MCP status line selection and authoring through `/statusline` when `tui_extensions_statusline` is enabled. (#61979, #61988, #62037) ## Bug Fixes - Preserved drafts, queued follow-ups, and interrupted input with their original threads when switching or resuming sessions. (#61216) - Kept permission profiles intact when changing working directories. (#61195) - Prevented late completion events from cancelling approvals for a newer turn. (#61801) - Preserved earlier edits in aggregate diffs when steering an active turn. (#61820) - Fixed scrolling that inserted stray characters, composer caret jumps, and redraws using stale terminal dimensions. (#62139, #62174, #62175) - Fixed enterprise MCP sign-in for identity providers that issue ID tokens without refresh tokens. (#62129) ## Chores - Updated source builds and release tooling to Rust 1.99.1. (#61205) - Updated XML dependencies and removed obsolete security advisory exceptions. (#62068) ## Changelog Full Changelog: https://github.com/openai/codex/compare/rust-v0.170.0...rust-v0.171.0 - #61178 Enable conversation-scoped speed selection for Cloud turns @bc-openai - #61180 Reclaim abandoned trusted Git repositories @etraut-openai - #61181 Fix pagination for unindexed threads sorted by ascending recency @etraut-openai - #61182 Support renaming cloud threads @bc-openai - #61183 Fix fork parent metadata in reconstructed TUI sessions @fcoury-oai - #61185 Defer Guardian context limits to request preparation @ioanbudea-OAI - #61186 Keep plan item IDs unique after completion @neerajrj-oai - #61187 Add authenticated GCS caching for Bazel CI and releases @dkovalenko-oai - #61194 Remove redundant legacy permission helpers @ioanbudea-OAI - #61195 Preserve permission profiles across working directory updates @ioanbudea-OAI - #61196 Clear inline image previews before history insertion can scroll them @dob-oai - #61197 Clarify remote message board and clock documentation @anp-oai - #61198 Add graceful agent-tree suspension and recovery @bryanashley - #61199 Align rollout trace replay with the same-binary contract @cassirer-openai - #61200 Unify TUI status content and layout caching in `StatusLine` @fcoury-oai - #61201 Clean up Rust error handling and OTLP test assertions @ioanbudea-OAI - #61205 Upgrade the Rust toolchain to 1.99.1 @ioanbudea-OAI - #61206 Enforce Linux seccomp restrictions for x32 syscalls @dhavalkapil-oai - #61207 Add SPIFFE JWT-SVID support for workload identity authentication @cooper-oai - #61210 Retain TUI delegation prompts in thread model history @won-openai - #61211 Add local status line provider configuration @joesavona - #61212 Add persistence support for status line provider selections @joesavona - #61213 Support model catalog Apps usage instructions @goodman-oai - #61215 Track dictation usage from accepted user inputs @chrisdong-oai - #61216 Preserve thread input across TUI session resumes @bc-openai - #61218 Hide attached threads from the agents overview @fcoury-oai - #61219 Hide TUI status details for read-only history @joesavona - #61220 Deduplicate overlapping Apps startup tool listings @charliemarsh-oai - #61221 Forward repeated signals through the npm wrapper @etraut-openai - #61222 Store rollout trace payloads in compressed shards @cassirer-openai - #61224 Include AppX event IDs in Windows runtime registration errors @zm-oai - #61227 Include Win32 error codes in MXC process launch failures @zm-oai - #61228 Add a missing semicolon in the realtime call test helper @bc-openai - #61248 Add MCP inventory support for statusline providers @joesavona - #61249 Resolve MCP statusline providers and read thread-scoped resources @joesavona - #61250 Add bounded streaming discovery for statusline providers @joesavona - #61251 Refactor history and notes tools around backend traits @euroelessar - #61253 Move rollout trace recording to a background worker @cassirer-openai - #61254 Fix Bazel tool resolution for bubblewrap and voice builds @bencodes-oai - #61255 Honor `PIP_INDEX_URL` for V8 Python dependencies @bencodes-oai - #61256 Block `request_key` in Linux network sandbox filters @dhavalkapil-oai - #61257 Preserve MCP registration scope across client operations @sayan-oai - #61259 Update realtime model and honor provider-specific voice defaults @cooper-oai - #61260 Accept ERS v1 and v2 tunnel JWTs @dines-oai - #61263 Refactor session window access and async compaction test setup @maxj-oai - #61264 Enable gzip and zstd support in the shared HTTP client @joeyp-openai - #61272 Set the environment field in statusline MCP requests @anp-oai - #61276 Consolidate Windows filesystem helper sandbox preparation @viyatb-oai - #61283 Accept explicit contribution context for MCP runtime configuration @sayan-oai - #61286 Add and preserve the marketplace auto-update setting @smurching-oai - #61337 Centralize managed network proxy startup @viyatb-oai - #61338 Record host-confirmed Code Mode wait targets in rollout traces @tz-openai - #61339 Add session-scoped stores for stable model-facing IDs @anp-oai - #61341 Add opt-in prompt image sanitization @vivi - #61342 Cancel abandoned cloud thread list requests @bc-openai - #61343 Title saved side chats from their first direct input @fcoury-oai - #61344 Add per-item retention policies for injected history @vivi - #61345 Add opt-in elastic budgeting for asynchronous Guardian requests @ioanbudea-OAI - #61346 Apply elastic entry caps to Guardian classifier evidence @ioanbudea-OAI - #61347 Reserve item ID overhead before elastic context allocation @ioanbudea-OAI - #61349 Add managed CA defaults for Deno and Google Cloud SDK @viyatb-oai - #61350 Reflect accepted cloud model selections in TUI status @bc-openai - #61351 Expose canonical network proxy policy patterns @viyatb-oai - #61352 Mark the TUI workspace as stable and enable it by default @fcoury-oai - #61353 Centralize TUI read-only view checks @bc-openai - #61354 Avoid rewriting unchanged Guardian channel descriptions @felixxia-oai - #61355 Upgrade Bazel rules_rs to 0.2.3 @seanh-oai - #61356 Add deferred MCP reprojection for the next regular turn @kchainani-oai - #61357 Clarify removal scope for profile MCP servers @etraut-openai - #61358 Add structured error kinds for exec-server process starts @andrewgu-oai - #61359 Resolve queued attachment paths against the thread working directory @etraut-openai - #61365 Limit MCP status discovery to the requested page @charliemarsh-oai - #61369 Use native permission profiles for diagnostics and sandbox checks @ioanbudea-OAI - #61370 Remove BuildBuddy presets from the checked-in Bazel configuration @hintz-openai - #61371 Add opt-in signer-service support for macOS releases @tarunoai - #61372 Add opt-in transparent TCP routing for Linux sandboxes @viyatb-oai - #61374 Delegate installed plugin recommendation filtering to the endpoint @zamoshchin-openai - #61379 Remove unused workspace entries and simplify Rust code @andrewgu-oai - #61380 Add cell-aware mouse selection to transcript tables @fcoury-oai - #61381 Update the harness provenance comment in context history @vivi - #61382 Make clock reads synchronous and remove client clock requests @euroelessar - #61383 Keep the composer responsive while opening new conversations @bc-openai - #61384 Support Orbit originators in model catalog partitioning @bromano-oai - #61385 Add configurable idle heartbeats to Code Mode gRPC streams @surajs-oai - #61386 Reuse cloud plugin catalogs across turns until invalidated @TAFOYA-OAI - #61387 Refresh plugin recommendations when tooling state changes @zamoshchin-openai - #61388 Include the Windows sandbox service in release builds @zm-oai - #61389 Reduce session submission channel memory with boxed payloads @charliemarsh-oai - #61390 Preserve active additional context across compaction @dylan-hurd-oai - #61391 Normalize managed network requirements before merging @viyatb-oai - #61393 Add optional release IDs to cloud plugin source locations @zamoshchin-openai - #61394 Add opt-in file-backed tool audio @hung-oai - #61395 Remove legacy sandbox-policy accessors from configuration @ioanbudea-OAI - #61397 Add default model provider selection to the app server @abhinav-oai - #61398 Restore shared connector metadata for Codex Apps tools @mzeng-openai - #61400 Retry incomplete cloud catalogs with exponential backoff @TAFOYA-OAI - #61401 Release MCP OAuth read locks before keyring access @celia-oai - #61402 Update Daybreak links in TUI cybersecurity notices @jamy-OAI - #61403 Add a structured status-line snapshot for the displayed thread @joesavona - #61405 Show the primary executor's workspace in TUI status surfaces @joesavona - #61408 Expose token usage totals in statusline state @joesavona - #61409 Expose context usage in statusline state @joesavona - #61410 Expose thread costs in statusline state and reject stale usage reads @joesavona - #61411 Ignore stale workspace headline requests and archived messages @joesavona - #61415 Expose the workspace headline in structured status-line state @joesavona - #61417 Expose Codex rate limits in statusline state @joesavona - #61418 Require a government workspace for the government backend @jackz100 - #61419 Resolve pasted Windows paths with wslpath in WSL @etraut-openai - #61420 Clear the user endpoint override during GCP setup @alexsong-oai - #61421 Use thread-scoped model catalogs for model lookups @abhinav-oai - #61422 Add keyboard table selection to the transcript @fcoury-oai - #61424 Refill the transcript viewport when collapsing tool groups @fcoury-oai - #61426 Enable incremental tools by default for Amazon Bedrock @akira-oai - #61428 Require OAuth CONNECT metadata for Portals TCP tunnels @apanasenko-oai - #61429 Refresh cloud plugin and skill catalogs during step preparation @TAFOYA-OAI - #61430 Instrument MCP OAuth credential store locks @celia-oai - #61434 Close abandoned tool dispatch traces as aborted @maxj-oai - #61437 Avoid starting dormant REPL servers for cleanup hooks @dob-oai - #61438 Use the provider default RPC in the TUI @abhinav-oai - #61444 Add process-local launch directories for shared MCP servers @charliemarsh-oai - #61446 Add thread ID inclusion filters to `thread/list` @btraut-openai - #61449 Add explicit cloud catalog refresh before turn admission @TAFOYA-OAI - #61452 Add optional tool output rendering for context window hints @euroelessar - #61461 Refactor session mail recording and compaction metadata capture @euroelessar - #61465 Move JSON byte counting into `codex-utils-string` @euroelessar - #61473 Persist timestamped realtime frames before acknowledging completion @zahan-oai - #61480 Add snapshot pagination for persisted realtime history @zahan-oai - #61481 Allow Ctrl+C to quit during Cloud session switches @bc-openai - #61486 Keep archived threads read-only and support unarchiving in the TUI @bc-openai - #61488 Harden Git remote collection and limit remote URL metadata @ningyi-oai - #61490 Preserve the current Cloud task until navigation is ready @bc-openai - #61494 Describe context and permissions per environment behind a feature flag @sayan-oai - #61535 Add trusted request-country metadata to Responses requests @goodman-oai - #61562 Unify model request hook lifecycles across sampling and compaction @tz-openai - #61577 Add response processing after completion and usage accounting @tz-openai - #61578 Recheck rollout budgets before starting tool work @tz-openai - #61579 Support fractional rollout budgets and configurable context reminders @tz-openai - #61580 Account for synchronous MCP post-hook costs in rollout budgets @tz-openai - #61581 Record neutral interruption markers for rollout budget stops @tz-openai - #61582 Gate new rollout work on shared budget exhaustion @tz-openai - #61583 Refactor rollout budget checks and turn completion state @tz-openai - #61587 Add request IDs and attempt numbers to model request hooks @tz-openai - #61595 Scope TUI history item lookups to their owning turns @etraut-openai - #61596 Keep agent overview refresh state scoped to each backend @bc-openai - #61599 Report Cloud stream recovery failures and cancel abandoned reads @bc-openai - #61600 Enable automatic titles for Cloud conversations @bc-openai - #61619 Return to the command center while archiving a task @bc-openai - #61622 Clarify window ID discovery in history tool guidance @euroelessar - #61663 Fix config reads with relative working directories @charliemarsh-oai - #61686 Defer plugin manifest loading until marketplace entries are selected @charliemarsh-oai - #61688 Add TUI review ownership and risk guidance @bc-openai - #61690 Keep pasted text inline in TUI goals @etraut-openai - #61692 Fix status detail truncation for wide Unicode characters @etraut-openai - #61694 Discard unsaved experimental feature edits on cancel @etraut-openai - #61697 Restore the terminal's original alternate-scroll mode @etraut-openai - #61706 Refactor exec to use the shared app-server client @etraut-openai - #61707 Support remote app servers in `codex exec` @etraut-openai - #61709 Avoid legacy transcript reads in the agents overview @etraut-openai - #61710 Enable slash-command completion for unavailable threads @etraut-openai - #61711 Open Windows desktop sessions through the registered URL handler @etraut-openai - #61713 Restore raw mode after returning from external editors @etraut-openai - #61724 Use `clip.exe` for single-line ASCII clipboard copies in WSL @etraut-openai - #61725 Fix Unicode grapheme highlighting in history search @etraut-openai - #61726 Preserve graphemes in mention search match highlighting @etraut-openai - #61727 Optimize default and attached thread listings with targeted indexes @dkovalenko-oai - #61730 Fix thread listing migration numbering and test setup @charliemarsh-oai - #61731 Keep bidirectional TUI headers stable during shimmer @etraut-openai - #61737 Preserve agent message phases in the cloud backend @bc-openai - #61738 Reuse matching thread inventories for lightweight MCP status pages @charliemarsh-oai - #61742 Mark cloud threads with a cloud icon in the TUI @bc-openai - #61744 Clarify shell-specific login guidance for `exec_command` @zm-oai - #61747 Refactor history and notes tool registration @euroelessar - #61762 Fix the missing label style in the mention truncation test @tz-openai - #61764 Simplify history and notes tool registration @euroelessar - #61771 Remove numeric positions from project responses @apanasenko-oai - #61784 Centralize permission profile compatibility helpers in protocol @ioanbudea-OAI - #61785 Strengthen workspace-write configuration tests @ioanbudea-OAI - #61795 Add lifecycle-managed Guardian housekeeping @felixxia-oai - #61797 Wire Guardian V3 housekeeping into the app server @felixxia-oai - #61798 Add a dedicated Guardian v3 housekeeper inference client @felixxia-oai - #61799 Route Guardian housekeeping through its dedicated sampler @felixxia-oai - #61800 Clarify managed human command checkpoint stability @felixxia-oai - #61801 Preserve active turn approvals when an earlier turn finishes late @jif-oai - #61817 Isolate app-server turn metadata by turn ID @jif-oai - #61820 Preserve aggregate turn diffs across steering @jif-oai - #61833 Avoid unnecessary rollout item copies during child thread creation @charliemarsh-oai - #61837 Avoid redundant store reads for loaded thread metadata @charliemarsh-oai - #61840 Resolve fork ancestry lazily for descending history pages @charliemarsh-oai - #61841 Close late command approval items when turns are interrupted @jif-oai - #61845 Defer code-mode tool catalog construction until first access @charliemarsh-oai - #61846 Reuse code-mode tool catalogs across executions @charliemarsh-oai - #61851 Allow network policy to enable transparent TCP capture @viyatb-oai - #61852 Broadcast read-state changes to experimental clients @yiqian-oai - #61854 Allow model metadata to control multi-agent tools in code mode @jif-oai - #61860 Add opt-in tool-call attribution for upstream proxies @hbuyle-oai - #61865 Report remote plugin local versions from matching cached packages @charliemarsh-oai - #61866 Preserve SQLite row metadata in thread listings @charliemarsh-oai - #61871 Preserve plugin skill picker ranking across version changes @bc-openai - #61878 Support flat subgraphs in native Mermaid flowcharts @etraut-openai - #61880 Preserve local Markdown document targets in transcript rendering @fcoury-oai - #61887 Add app-server file preview read helpers @fcoury-oai - #61888 Add read-only document previews to the TUI sidebar @fcoury-oai - #61889 Add search, selection, and link navigation to document previews @fcoury-oai - #61890 Open source file links at line and column targets in the TUI @fcoury-oai - #61894 Acknowledge agent messages and clarify how to wake idle recipients @rjmarsan - #61896 Bound rollout trace capture with configurable storage limits @cassirer-openai - #61897 Add opt-in UTC dates to environment context @sayan-oai - #61898 Tie Linux sandbox proxy bridges to the launcher lifetime @viyatb-oai - #61900 Fix proc mount fallback detection for Bubblewrap 0.12 @viyatb-oai - #61902 Filter thread listings before scanning rollout previews @charliemarsh-oai - #61908 Pass a per-call voice session ID through TUI realtime startup @chrisdong-oai - #61909 Add opt-in item-boundary interrupts for subagents @jif-oai - #61910 Preserve native sandbox executable paths and PTY dispatch names @viyatb-oai - #61911 Default Statsig metrics exports to Zstandard compression @tvh-openai - #61915 Fall back to the built-in status line for invalid providers @joesavona - #61916 Preserve remote compaction summaries when interrupting child agents @jif-oai - #61919 Add sidebar image previews and present generated images after answers @fcoury-oai - #61920 Track conversation files for the TUI sidebar @fcoury-oai - #61922 Add conversation files to the TUI sidebar chooser @fcoury-oai - #61926 Allow agent search by thread ID and Git branch @bc-openai - #61928 Encapsulate app-server thread-list locking in `ThreadListState` @charliemarsh-oai - #61929 Freeze Guardian review views at ordered evidence cutoffs @felixxia-oai - #61931 Log Windows MXC deny-read glob expansion timings @iceweasel-oai - #61934 Allow independent app-server thread resumes to overlap @charliemarsh-oai - #61939 Preserve partial Git metadata when worktree status is slow @ningyi-oai - #61940 Add saved environment listing to the Cloud client @bc-openai - #61941 Route turn-start client metadata through a dedicated helper @chrisdong-oai - #61942 Centralize exec-server launches in `WrappedSandboxCommand` @viyatb-oai - #61945 Allow host-started forks to inherit the root session ID @rhan-oai - #61946 Include enclosing tool context in nested Guardian approvals @ankushg - #61947 Mask system SSH config drop-ins in the Linux sandbox @viyatb-oai - #61949 Extract user input preparation into `codex-user-input` @anp-oai - #61951 Refactor composer undo and redo history @etraut-openai - #61954 Add undo and redo to the non-Vim prompt composer @etraut-openai - #61955 Bound item loading for paginated `read_thread` calls @charliemarsh-oai - #61956 Support saved environments when starting Cloud threads @bc-openai - #61958 Support turn-scoped exhaustion in host usage accounting @rhan-oai - #61959 Prewarm restored file images on cold resume @bkotsopoulos -