mirror of
https://github.com/openai/codex.git
synced 2026-10-22 17:58:46 +00:00
Ian MacLeod
f0a1b8f084
- Added experimental `/voice` conversations with live transcripts and microphone controls on supported builds, enabled through `/experimental`. (#43581, #43651, #44331) - The TUI now shows live reasoning summaries in the status row and completion timestamps after successful turns. (#43558, #43921) - Added task hiding, archiving, and deletion in the agents overview, plus worktree ownership details and confirmed deletion of clean managed worktrees. (#43942, #44424, #44433) - Added Touch ID verification for MCP requests in local TUI sessions on supported Macs. (#43624, #43712, #43715) - Added configurable daemon update schedules and `codex app-server daemon update`; saved threads and active goals can recover after daemon restarts. (#43542, #43562, #44314) - Amazon Bedrock can now obtain AWS credentials from configured commands, with caching, expiration-based refresh, and authentication recovery. (#44028) ## Bug Fixes - Accepted prompts are now saved even when compaction fails before a turn starts. (#44487) - Fixed missed tmux resizes, transcript viewport restoration, and stale history appearing after switching threads. (#43603, #43889, #43994) - MCP servers now report expired OAuth credentials accurately and provide reconnect guidance when token refresh fails. (#43947, #44359) - Automatic approval reviews now preserve complete actions and authorization evidence more reliably, retry transient failures, and distinguish review failures from unsafe-action findings. (#44482, #44569, #44570) - Switching accounts now invalidates remote-control sessions, cached WebSocket state, and model catalogs belonging to the previous identity. (#43906, #44341, #44489) - Blocked Windows-process escapes from restricted WSL sandboxes and hardened brokered shell snapshots against credential exposure. (#44286, #43909, #44040) ## Chores - Aligned Python SDK and runtime publishing with stable CLI releases, using matching versions and verifying runtime assets before SDK publication. (#44067) ## Changelog Full Changelog: https://github.com/openai/codex/compare/rust-v0.154.0...rust-v0.155.0 - #43521 Preserve standalone release pins during daemon updates @copyberry - #43524 Restrict MCP user verification and add workspace-scoped identity @copyberry - #43527 Discount an approval's own code-mode wrapper from Guardian score lag @copyberry - #43529 Ensure the standalone updater runs on managed daemon starts @copyberry - #43533 Fix jemalloc tools and compiler flags for Bazel musl builds @copyberry - #43534 Preserve Guardian context sections and share planned-action rendering @copyberry - #43538 Move Guardian permission context into the shared section registry @copyberry - #43540 Preserve the multi-agent version when forking at a turn cutoff @copyberry - #43542 Make app-server daemon automatic updates configurable @copyberry - #43545 Preserve fork runtime versions without loading full model context @copyberry - #43547 Add user-verification provider abstractions and RPC adapters @copyberry - #43552 Record the launched app-server executable identity in PID files @copyberry - #43558 Show completion timestamps after successful TUI turns @copyberry - #43562 Add an explicit app-server daemon update command @copyberry - #43568 Wire app-server user verification RPCs to the native provider @copyberry - #43570 Manage synchronous Guardian reviewers through the thread manager @copyberry - #43572 Make the managed app-server shutdown grace period configurable @copyberry - #43576 Group adjacent computer actions in the TUI @copyberry - #43581 Add live WebRTC voice conversations to the TUI @copyberry - #43584 Remove a stale transcript field assignment from the TUI @copyberry - #43595 Centralize bounded Guardian review evidence in guardian-context @copyberry - #43597 Move trusted tool metadata into shared Guardian context @copyberry - #43599 Move trusted skill evidence into the Guardian context registry @copyberry - #43601 Move Guardian image selection into shared context sections @copyberry - #43602 Move Guardian REPL evidence rendering into the shared context registry @copyberry - #43603 Recover missed tmux resize notifications in the TUI @copyberry - #43604 Exclude base instructions from the bundled model catalog @copyberry - #43619 Add a stable TUI/app-server version comparison helper @copyberry - #43621 Add worktree classification to thread telemetry @copyberry - #43622 Warn when the connected Codex service is older than the CLI @copyberry - #43624 Add macOS user verification with Secure Enclave signing @copyberry - #43645 Expand TUI regression coverage for realtime voice conversations @copyberry - #43651 Add voice mute shortcut and recording activity indicators @copyberry - #43656 Animate live voice transcripts with split-flap tiles @copyberry - #43676 Style spoken prompts and link workspace files in voice transcripts @copyberry - #43683 Move voice controls into a dedicated composer strip @copyberry - #43690 Make the voice mute shortcut configurable in the TUI @copyberry - #43695 Stabilize realtime voice meter sampling across redraws @copyberry - #43698 Make older app-server notices configurable in the TUI @copyberry - #43699 Preserve split-flap animation state when voice transcripts scroll @copyberry - #43702 Add a TUI user verification prompt component @copyberry - #43704 Disable clock synchronization in the voice audio sink @copyberry - #43708 Add TUI request bookkeeping for user verification @copyberry - #43712 Enable MCP user verification in the TUI @copyberry - #43715 Enable user verification for the bundled TUI on supported devices @copyberry - #43790 Limit app-server storage metrics to session directories @copyberry - #43795 Pin request reasoning effort while configuration overrides are active @copyberry - #43796 Preserve reasoning effort through compaction and reset it on success @copyberry - #43797 Add configurable memory versions with isolated storage @copyberry - #43798 Batch non-user history eviction to preserve Guardian transcript deltas @copyberry - #43799 Prioritize human evidence in memory v2 extraction @copyberry - #43800 Add summary-only extraction for memory v2 @copyberry - #43805 Centralize Guardian context composition @copyberry - #43806 Centralize Guardian transcript policy in context profiles @copyberry - #43808 Move v2 extraction chunking into the memory writer @copyberry - #43813 Add dedicated memory v2 consolidation and read prompts @copyberry - #43827 Add memory dual writing and v2 readiness reporting @copyberry - #43842 Wait for parent idle before rollback in guardian fork tests @copyberry - #43844 Add staged enterprise OIDC login and coordinated logout @copyberry - #43846 Include completed commentary in the `/copy` picker @copyberry - #43848 Preserve runtime workspace roots across thread resume @copyberry - #43853 Add canonical permission translation for MXC execution requests @copyberry - #43870 Close MCP stderr readers on client teardown @copyberry - #43873 Handle undefined values before JSON serialization in code mode @copyberry - #43876 Detach Unix hook commands from the controlling terminal @copyberry - #43884 Close active network proxy connections on teardown @copyberry - #43889 Fix transcript viewer restoration and half-page scrolling @copyberry - #43894 Preserve thread identity in code-mode tool dispatch traces @copyberry - #43895 Preserve `__oailb` routing cookies in ChatGPT HTTP clients @copyberry - #43897 Persist provider and auth identity with model catalog caches @copyberry - #43900 Propagate Apps tool refreshes to existing threads @copyberry - #43903 Extract Windows deny-read glob scan planning into protocol @copyberry - #43906 Scope model catalog caches to the current provider and auth identity @copyberry - #43907 Preserve complete shell snapshot exports through filtering and replay @copyberry - #43909 Protect shell snapshots when credential brokerage is enabled @copyberry - #43912 Keep Guardian reviewers on summary-based compaction @copyberry - #43913 Add tracing for project instructions and filesystem sandbox operations @copyberry - #43915 Remove the repository devcontainer configurations @copyberry - #43918 Expose the queued event count on `CodexThread` @copyberry - #43921 Show streaming reasoning summaries in the TUI status row @copyberry - #43925 Add cancellation for native user-verification RPCs @copyberry - #43927 Rename thread artifacts to attachments in the state database @copyberry - #43930 Avoid Windows sandbox setup for irrelevant proxy port changes @copyberry - #43934 Track voice session lifecycle metrics in the TUI @copyberry - #43936 Stabilize subagent and unified exec test fixtures @copyberry - #43937 Tag TUI startup metrics with terminal and multiplexer categories @copyberry - #43939 Add executor-context filesystem permission helpers @copyberry - #43942 Show worktree owner details and add confirmed deletion @copyberry - #43943 Gate new turn submissions on host shutdown admission @copyberry - #43947 Surface MCP reconnect signals when expired OAuth tokens cannot refresh @copyberry - #43948 Show configured app-server updater settings in doctor @copyberry - #43949 Add transactional thread attachment mutations to the state runtime @copyberry - #43950 Keep app-server thread RPCs active until delegated work completes @copyberry - #43953 Preserve per-image generation IDs in image generation analytics @copyberry - #43954 Cache protected shell snapshots and harden capture cleanup @copyberry - #43956 Increase the TUI thread capability test stack to 12 MiB @copyberry - #43959 Gate new app-server work during graceful shutdown @copyberry - #43961 Use curly apostrophes in protocol error messages @copyberry - #43966 Heap-allocate the resume future in the legacy history test @copyberry - #43983 Bundle signed voice resources in macOS releases @copyberry - #43994 Clear stale transcript history when switching threads @copyberry - #44002 Encapsulate executed tool call metadata recording @copyberry - #44025 Refresh workspace lockfile before building macOS voice releases @copyberry - #44027 Support image attachments in agents overview background tasks @copyberry - #44028 Add AWS credential export commands for Amazon Bedrock @copyberry - #44032 Generate Python SDK types from repository app-server schemas @copyberry - #44038 Handle copied credentials in the broker and shell snapshots @copyberry - #44040 Harden credential handling in shell snapshots and replay @copyberry - #44043 Use `StartThreadOptions` across thread fork APIs @copyberry - #44049 Extract credential broker environment and registry helpers @copyberry - #44053 Test Python SDK against the built CLI and installed runtime @copyberry - #44055 Gate Python SDK publishing on runtime availability and verify PyPI files @copyberry - #44056 Add configurable credential providers to the network proxy @copyberry - #44060 Raise Guardian's action review limit to 200,000 bytes @copyberry - #44061 Build Python SDK artifacts before publishing the runtime @copyberry - #44062 Fix voice runtime release builds and packaging @copyberry - #44066 Extend configured credential brokerage to embedded aliases @copyberry - #44067 Publish Python packages after stable CLI releases @copyberry - #44068 Preserve credential broker destinations across environment filtering @copyberry - #44070 Reject empty audio payloads in data URLs @copyberry - #44072 Support configured credential providers across shell snapshots @copyberry - #44077 Refactor credential-broker tunnel protocol detection @copyberry - #44080 Make staged macOS voice runtimes writable before packaging @copyberry - #44084 Expose Python SDK history selection and per-turn options @copyberry - #44086 Add untrusted external messages to the Python SDK @copyberry - #44089 Support credential brokering in plaintext HTTP tunnels @copyberry - #44101 Handle empty voice arguments in macOS release packaging @copyberry - #44121 Reuse MCP bindings while cached servers remain dormant @copyberry - #44138 Coordinate rollout compression with active thread writers @copyberry - #44164 Add Guardian context cost and request token telemetry @copyberry - #44166 Add aggregate budget enforcement for Guardian context @copyberry - #44175 Attach compressed rollouts to diagnostic reports as JSONL @copyberry - #44176 Reduce TUI stack usage during session transitions @copyberry - #44181 Use explicit histogram buckets for Guardian context metrics @copyberry - #44183 Release persistent writers when session startup is cancelled @copyberry - #44198 Preserve voice indicator styles during composer sparkle effects @copyberry - #44200 Use captured step settings when building model context @copyberry - #44202 Use captured step model settings for extension context @copyberry - #44207 Measure total exec-server request duration including queueing @copyberry - #44226 Continue rollout searches when a compressed rollout cannot be searched @copyberry - #44227 Extract Guardian review policy into a dedicated crate @copyberry - #44238 Fix MCP elicitation cancellation and reset state on reconnect @copyberry - #44241 Handle credential provider source remapping across config layers @copyberry - #44242 Use captured step settings for tool planning and execution @copyberry - #44243 Use the originating model when recording conversation history @copyberry - #44248 Preserve tool output truncation budgets across resume and fork @copyberry - #44249 Normalize image detail for the receiving model @copyberry - #44250 Remove retired model entries while preserving migration prompts @copyberry - #44252 Move synchronous Guardian orchestration into the reviewer extension @copyberry - #44255 Always use streamed remote compaction for supported providers @copyberry - #44259 Remove the Windows `/sandbox-add-read-dir` slash command @copyberry - #44273 Remove the unused legacy remote compaction implementation @copyberry - #44276 Avoid duplicate reasoning effort updates during turn recovery @copyberry - #44277 Surface environment startup failure reasons to the model @copyberry - #44281 Enforce complete request budgets for Guardian reviews @copyberry - #44283 Persist loaded threads before managed daemon shutdown @copyberry - #44284 Add telemetry for the Windows system config namespace @copyberry - #44285 Preserve prewarmed reasoning effort across replay and early rollback @copyberry - #44286 Block WSL interop escapes from restricted filesystem sandboxes @copyberry - #44288 Prevent command hooks from hanging on blocked stdin @copyberry - #44289 Expand MXC volume grants and resolve deny globs @copyberry - #44290 Allow user-requested goal pauses through `update_goal` @copyberry - #44293 Enforce the async Guardian classifier's complete input budget @copyberry - #44297 Isolate the hook pipe I/O timeout test from shell startup files @copyberry - #44298 Set turn triggers for guardian and memory requests @copyberry - #44299 Record thread recovery candidates on managed daemon shutdown @copyberry - #44307 Add opt-in provisioned macOS CLI release candidates @copyberry - #44311 Honor shared Retry-After deadlines for remote control @copyberry - #44314 Restore saved threads when the managed daemon restarts @copyberry - #44318 Give hosted Codex Apps an independent MCP protocol opt-in @copyberry - #44320 Block goals after three empty automatic continuation turns @copyberry - #44325 Return the prompt hash in upload responses @copyberry - #44327 Prevent filesystem-root read denies in the Windows sandbox @copyberry - #44328 Clear pending TUI questions when accepting a new prompt @copyberry - #44330 Add paginated thread attachment listing to the state runtime @copyberry - #44331 Expose voice conversations in experimental features @copyberry - #44332 Persist disabled plugin IDs in thread settings @copyberry - #44336 Add bounded tool-result metadata support to executed tool calls @copyberry - #44337 Return to the agent command center after archiving on shared servers @copyberry - #44341 Bind remote-control sessions to their authentication owner @copyberry - #44344 Open tasks with Right from the agents overview @copyberry - #44346 Support native verification in MCP tool continuations @copyberry - #44349 Distinguish forked sessions in session-start hooks @copyberry - #44350 Add thread attachment operations with coordinated deletion @copyberry - #44352 Remove path-bearing fields from Guardian review analytics @copyberry - #44354 Extract shared footer hint wrapping in the TUI @copyberry - #44359 Report OAuth authentication failures in MCP status snapshots @copyberry - #44360 Return focus to the agents overview composer on Escape @copyberry - #44377 Update the forked-thread hook test to use `StartThreadOptions` @copyberry - #44392 Add opt-in model discovery for OpenAI API keys @copyberry - #44400 Start Python SDK turn subscriptions at their attachment point @copyberry - #44424 Add a hide shortcut to the agents overview @copyberry - #44433 Add archive and delete actions to the agents overview @copyberry - #44472 Harden Code Mode tool-call completeness tracking @copyberry - #44482 Improve Guardian retries and review failure reporting @copyberry - #44487 Preserve incoming prompts when pre-turn compaction fails @copyberry - #44489 Reset cached WebSocket state when auth ownership changes @copyberry - #44492 Distinguish HTTP quota errors from rate limits @copyberry - #44493 Bound MCP descriptions separately from Guardian action JSON @copyberry - #44521 Decouple session isolation from subagent attribution @copyberry - #44523 Bound app-server stdio shutdown and handle Unix SIGTERM gracefully @copyberry - #44536 Move Guardian reviewer settings and execution into the reviewer crate @copyberry - #44544 Move Guardian reporting and denial accounting into the extension @copyberry - #44548 Add MIME-filtered resource listing for Codex Apps @copyberry - #44564 Add app-server APIs for stored thread attachments @copyberry - #44569 Preserve complete actions in Guardian approval reviews @copyberry - #44570 Preserve Guardian authorization evidence until request budgeting @copyberry - #44571 Allow extensions to select MCP protocol mode per HTTP server @copyberry - #44574 Use captured action settings for Guardian reviews @copyberry - #44575 Tie network approval reviews to their originating execution @copyberry - #44580 Support symbolic `:root` filesystem policies in MXC @copyberry - #44586 Remove `repo_url` from skill invocation analytics events @copyberry
Codex CLI is a coding agent from OpenAI that runs locally on your computer.
If you want Codex in your code editor (VS Code, Cursor, Windsurf), install in your IDE.
If you want the desktop app experience, run
codex app or visit the Codex App page.
If you are looking for the cloud-based agent from OpenAI, Codex Web, go to chatgpt.com/codex.
Quickstart
Installing and running Codex CLI
Run the following on Mac or Linux to install Codex CLI:
curl -fsSL https://chatgpt.com/codex/install.sh | sh
Run the following on Windows to install Codex CLI:
powershell -ExecutionPolicy ByPass -c "irm https://chatgpt.com/codex/install.ps1 | iex"
The standalone installers download from https://releases.openai.com/codex by default and fall back to GitHub Releases if a metadata or asset download is unavailable. To force GitHub Releases, set CODEX_INSTALLER_USE_RELEASES_OPENAI_COM to false (0 and no are also accepted):
curl -fsSL https://chatgpt.com/codex/install.sh | CODEX_INSTALLER_USE_RELEASES_OPENAI_COM=false sh
$env:CODEX_INSTALLER_USE_RELEASES_OPENAI_COM='false'; irm https://chatgpt.com/codex/install.ps1 | iex
Codex CLI can also be installed via the following package managers:
# Install using npm
npm install -g @openai/codex
# Install using Homebrew
brew install --cask codex
Then simply run codex to get started.
You can also go to the latest GitHub Release and download the appropriate binary for your platform.
Each GitHub Release contains many executables, but in practice, you likely want one of these:
- macOS
- Apple Silicon/arm64:
codex-aarch64-apple-darwin.tar.gz - x86_64 (older Mac hardware):
codex-x86_64-apple-darwin.tar.gz
- Apple Silicon/arm64:
- Linux
- x86_64:
codex-x86_64-unknown-linux-musl.tar.gz - arm64:
codex-aarch64-unknown-linux-musl.tar.gz
- x86_64:
Each archive contains a single entry with the platform baked into the name (e.g., codex-x86_64-unknown-linux-musl), so you likely want to rename it to codex after extracting it.
Using Codex with your ChatGPT plan
Run codex and select Sign in with ChatGPT. We recommend signing into your ChatGPT account to use Codex as part of your Plus, Pro, Business, Edu, or Enterprise plan. Learn more about what's included in your ChatGPT plan.
You can also use Codex with an API key, but this requires additional setup.
Docs
This repository is licensed under the Apache-2.0 License.