stdio servers now use a new client backed by the official rust MCP sdk
stdio servers can specify a cwd
All servers can specify enabled_tools or disabled_tools
Streamable HTTP servers can specify scopes during codex mcp login
Improved startup error messages
Better instruction following for calling tools
Configuration options for forced_login_method and forced_chatgpt_workspace_id which can be paired with managed configs to give enterprises more control over Codex usage.
You can now tag @codex on a teammate’s pull request to ask clarifying questions, request a follow-up, or ask Codex to make changes. GitHub Issues now also support @codex mentions, so you can kick off tasks from any issue, without leaving your workflow.
We are in the middle of significantly improving support for connecting to MCP servers including support for streamable http servers with optional bearer token or oauth login support.
Enable the experimental_use_rmcp_client = true flag in config.toml to use the new MCP stack. If you encounter issues, create a GitHub issue and add the mcp label. We would like to remove the experimental flag soon and will continue to expand MCP support over time.
What's Changed
[MCP] Fix the bearer token authorization header by @gpeal in #4846
Print codex resume note when quitting after codex resume by @nornagon-openai in #4695
Use Number instead of BigInt for TokenCountEvent by @ccy-oai in #4856
Codex is now generally available with 3 new features — @Codex in Slack, Codex SDK, and new admin tools.
@Codex in Slack
You can now questions and assign tasks to Codex directly from Slack. See the Slack guide to get started.
Codex SDK
Integrate the same agent that powers the Codex CLI inside your own tools and workflows with the Codex SDK in Typescript. With the new Codex GitHub Action, you can easily add Codex to CI/CD workflows. See the Codex SDK guide to get started.
import { Codex } from "@openai/codex-sdk";const agent = new Codex();const thread = await agent.startThread();const result = await thread.run("Explore this repo");console.log(result);const result2 = await thread.run("Propose changes");console.log(result2);
New admin controls and analytics
ChatGPT workspace admins can now edit or delete Codex Cloud environments. With managed config files, they can set safe defaults for CLI and IDE usage and monitor how Codex uses commands locally. New analytics dashboards help you track Codex usage and code review feedback. Learn more in the enterprise admin guide.
Availability and pricing updates
The Slack integration and Codex SDK are available to developers on ChatGPT Plus, Pro, Business, Edu, and Enterprise plans starting today, while the new admin features will be available to Business, Edu, and Enterprise.
Beginning October 20, Codex Cloud tasks will count toward your Codex usage. Review the Codex pricing guide for plan-specific details.
Codex CLI Release: 0.45.0
Click to reveal release details
To install this version of Codex CLI, run:
$ npm install -g @openai/codex@0.45.0
Breaking changes
codex login --api-key was replaced with codex login --with-api-key that reads the key from stdin to improve security.
New responses-api-proxy component to simplify and secure response handling/proxying (#4246)
Added secure mode: CODEX_SECURE_MODE=1 to restrict process observability (#4220)
More explicit codex exec events, including item.started and improved output display (#4177, #4250, #4113)
Safer command execution: dangerous command checks in core and Windows-specific safety (#4211, #4119)
UX improvements: revamped /status with clearer presentation and details, pageless session list, approvals moved to ListSelectionView, better MCP tool call styling, and hiding status indicator during streaming (#3194, #4275, #3871, #4101, #4196)
Stability and correctness: token usage fix for compaction, SSE mounting reliability, and improving bugs on editing previous conversations (#4281, #4264, #4237)
IDE extension reliability: fix login with API key when .codex directory is missing (#4258)
Internal cleanup and refactors for state, footer logic, and unused code (#4174, #4259, #4310)
Rate limits visibility: You can see when is your limits resetting (#4111) and see your usage when rate limited (#4102)
You can specify output schema in exec mode using output-schema. (#4079)
Ripgrep (rg) is now vendored into the npm release (#3660), which should fix issues such as #3842 where the postinstall step of @vscode/ripgrep would sometimes fail when installing codex via npm.
Merged PRs:
#4090 feat: readiness tool #4088 chore: compact do not modify instructions #4069 refactor codex card layout #4105 Use anyhow::Result in tests for error propagation #4079 Add exec output-schema parameter #4102 Send limits when getting rate limited #3461 [exec] add include-plan-tool flag and print it nicely #3660 fix: vendor ripgrep in the npm module #4112 fix: npm publish --tag alpha when building an alpha release #4031 chore(deps): bump serde from 1.0.224 to 1.0.226 in /codex-rs #4027 chore(deps): bump log from 0.4.27 to 0.4.28 in /codex-rs #4030 chore(deps): bump tempfile from 3.20.0 to 3.22.0 in /codex-rs #4135 nit: update auto compact to 250k #4137 nit: drop instruction override for auto-compact #4140 chore: extract readiness in a dedicated utils crate #4111 Add Reset in for rate limits #4156 nit: 350k tokens #4124 chore: upgrade to Rust 1.90 #4162 fix: add tolerance for ambiguous behavior in gh run list #4154 chore: remove once_cell dependency from multiple crates #2156 chore(deps): bump unicode-width from 0.1.14 to 0.2.1 in /codex-rs #4028 chore(deps): bump chrono from 0.4.41 to 0.4.42 in /codex-rs
GPT-5-Codex is now available in the Responses API, and you can also use it with your API Key in the Codex CLI.
We plan on regularly updating this model snapshot.
It is available at the same price as GPT-5. You can learn more about pricing and rate limits for this model on our model page.
This release includes only one change on top of 0.37.0, which is a fix to the npm-publish step in the release process (#3806). Updating npm with the latest release is now fully automated, and if you visit https://www.npmjs.com/package/@openai/codex#provenance-details-header, you should now see this special Provenance badge:
Merged PRs:
[#3806] fix: specify --repo when calling gh (#3806)
Mostly small quality-of-life fixes in this release.
Highlights
Updated the npm release process to use "trusted publishing" (#3431). Though apparently there was a bug, so we may not see the benefit until 0.38.0. Fix is in #3806.
Merged PRs
[#3329] notifications on approvals and turn end (#3329)
[#3659] chore: restore prerelease logic in rust-release.yml (#3659)
[#3617] chore(deps): bump serde_json from 1.0.143 to 1.0.145 in /codex-rs (#3617)
[#3664] chore: simplify dep so serde=1 in Cargo.toml (#3664)
[#3618] chore(deps): bump serde from 1.0.219 to 1.0.223 in /codex-rs (#3618)
[#3619] chore(deps): bump wildmatch from 2.4.0 to 2.5.0 in /codex-rs (#3619)
GPT-5-Codex is a version of GPT-5 further optimized for agentic coding in Codex.
It’s available in the IDE extension and CLI when you sign in with your ChatGPT account.
It also powers the cloud agent and Code Review in GitHub.
To learn more about GPT-5-Codex and how it performs compared to GPT-5 on software engineering tasks, see our announcement blog post.
Image outputs
When working in the cloud on front-end engineering tasks, GPT-5-Codex can now display screenshots of the UI in Codex web for you to review. With image output, you can iterate on the design without needing to check out the branch locally.
New in Codex CLI
You can now resume sessions where you left off with codex resume.
Context compaction automatically summarizes the session as it approaches the context window limit.
Breaking change:OPENAI_API_KEY is no longer read from the environment
API login is no longer implicit; we do not pick up OPENAI_API_KEY from the environment. To use an API key programmatically, run the following once:
codex login --api-key "your-api-key-here"
and this value will be stored in CODEX_HOME/auth.json.
The next time you run codex, the API key stored in auth.json will be used automatically.
To switch to ChatGPT login instead of using your API key, run codex login.
Note that codex logout will remove whatever login credential is stored in auth.json (API key or ChatGPT auth token).
Highlights
Introducing our newest model GPT-5-Codex
GPT-5-Codex works faster through easy tasks and harder on complex tasks, improves on code quality, and is more steerable with AGENTS.md. Learn More
Resuming old sessions
Resuming old conversations with codex resume (#3537, #3625).
Unified execution and reliability
Unified execution improvements (#3288, #3479) and race-condition fix (#3644); include command output on timeouts (#3576); do not execute when command parses as a patch (#3382); improved sandbox timeout handling (#3435); flaky test fixes (#3596, #3564).
Auth and login experience
login polish (#3632); fix issue allowing to be both logged in and using API key simultaneously (#3611); fix stray login URL characters (#3639); fix get_auth_status for custom providers (#3581).
JSON-RPC and MCP
JSON-RPC: SetDefaultModel and None clears default (#3512, #3529), UserInfo (#3428), reasoning_effort in NewConversationResponse (#3506); initial MCP interface and docs (#3543, #3507, #3345); standardized shell description (#3514).
TUI onboarding and polish
New onboarding flow and animation (#3398, #3631, #3590, #3627), consistent headers on resume (#3615, #3592), improved spacing/padding (#3469, #3471, #3472, #3474), clearer interruption and status styling (#3470), numerous UX nits addressed (#3650, #3534).
Core platform
Experimental automatic context compaction (#3446); expanded default sandbox (#3483); Azure Responses API workaround (#3528); preserve more item IDs in Azure (#3542); logging and clippy improvements (#3488, #3489); just test via nextest (#3508).
All merged PRs
fix: change MIN_ANIMATION_HEIGHT so show_animation is calculated correctly (#3656)
IDE extension (Compatible with VS Code, Cursor, Windsurf)
Codex now runs in your IDE with an interactive UI for fast local iteration. Easily switch between modes and reasoning efforts.
Sign in with ChatGPT (IDE & CLI)
One-click authentication that removes API keys and uses ChatGPT Enterprise credits.
Move work between local ↔ cloud
Hand off tasks to Codex web from the IDE with the ability to apply changes locally so you can delegate jobs without leaving your editor.
Code Reviews
Codex goes beyond static analysis. It checks a PR against its intent, reasons across the codebase and dependencies, and can run code to validate the behavior of changes.
Mid August update
Image inputs
You can now attach images to your prompts in Codex web. This is great for asking Codex to implement frontend changes or follow up on whiteboarding sessions.
Container caching
Codex now caches containers to start new tasks and followups 90% faster, dropping the median start time from 48 seconds to 5 seconds. You can optionally configure a maintenance script to update the environment from its cached state to prepare for new tasks. See the docs for more.
Automatic environment setup
Now, environments without manual setup scripts automatically run the standard installation commands for common package managers like yarn, pnpm, npm, go mod, gradle, pip, poetry, uv, and cargo. This reduces test failures for new environments by 40%.
Best of N
Codex can now generate multiple responses simultaneously for a single task, helping you quickly explore possible solutions to pick the best approach.
Fixes & improvements
Added some keyboard shortcuts and a page to explore them. Open it by pressing ⌘-/ on macOS and Ctrl+/ on other platforms.
Added a “branch” query parameter in addition to the existing “environment”, “prompt” and “tab=archived” parameters.
Added a loading indicator when downloading a repo during container setup.
Added support for cancelling tasks.
Fixed issues causing tasks to fail during setup.
Fixed issues running followups in environments where the setup script changes files that are gitignored.
Improved how the agent understands and reacts to network access restrictions.
Increased the update rate of text describing what Codex is doing.
Increased the limit for setup script duration to 20 minutes for Pro, Team, and Business users.
Polished code diffs: You can now option-click a code diff header to expand/collapse all of them.
June update
Agent internet access
Now you can give Codex access to the internet during task execution to install dependencies, upgrade packages, run tests that need external resources, and more.
Internet access is off by default. Plus, Pro, and Team users can enable it for specific environments, with granular control of which domains and HTTP methods Codex can access. Internet access for Enterprise users is coming soon.
Now you can update existing pull requests when following up on a task.
Voice dictation
Now you can dictate tasks to Codex.
Fixes & improvements
Added a link to this changelog from the profile menu.
Added support for binary files: When applying patches, all file operations are supported. When using PRs, only deleting or renaming binary files is supported for now.
Fixed an issue on iOS where follow up tasks where shown duplicated in the task list.
Fixed an issue on iOS where pull request statuses were out of date.
Fixed an issue with follow ups where the environments were incorrectly started with the state from the first turn, rather than the most recent state.
Fixed internationalization of task events and logs.
Improved error messages for setup scripts.
Increased the limit on task diffs from 1 MB to 5 MB.
Increased the limit for setup script duration from 5 to 10 minutes.
Polished GitHub connection flow.
Re-enabled Live Activities on iOS after resolving an issue with missed notifications.
Removed the mandatory two-factor authentication requirement for users using SSO or social logins.
Reworked environment page
It’s now easier and faster to set up code execution.
Fixes & improvements
Added a button to retry failed tasks
Added indicators to show that the agent runs without network access after setup
Added options to copy git patches after pushing a PR
Added support for unicode branch names
Fixed a bug where secrets were not piped to the setup script
Fixed creating branches when there’s a branch name conflict.
Fixed rendering diffs with multi-character emojis.
Improved error messages when starting tasks, running setup scripts, pushing PRs, or disconnected from GitHub to be more specific and indicate how to resolve the error.
Improved onboarding for teams.
Polished how new tasks look while loading.
Polished the followup composer.
Reduced GitHub disconnects by 90%.
Reduced PR creation latency by 35%.
Reduced tool call latency by 50%.
Reduced task completion latency by 20%.
Started setting page titles to task names so Codex tabs are easier to tell apart.
Tweaked the system prompt so that agent knows it’s working without network, and can suggest that the user set up dependencies.
Updated the docs.
Codex in the ChatGPT iOS app
Start tasks, view diffs, and push PRs—while you’re away from your desk.
