Codex changelog & version tracker
Continuously monitored from upstream releases, docs and issues. Routine releases publish automatically and the current version below is limited to a small allowlist of fields; everything else is reviewed by a human before it appears here.
Change timeline
- release
Python SDK 0.154.0 adds ExternalMessage, new reasoning-effort tiers, and history-selection options
The openai-codex Python SDK reached 0.154.0, matching CLI 0.154.0's runtime. New APIs let external content join or start a turn with tool-level authority, add 'max' and 'ultra' reasoning-effort values, and add include_turns/turn_service_tier/source options for resume and fork. Migration notes: HookMetadata now wraps its handler in .root (e.g. hook.root.command instead of hook.command), some notifications now have typed payloads instead of generic .params, and custom codex_bin overrides need CLI 0.151.0+ for these features.
Source ↗ - release
Codex CLI 0.154.0 stable released with worktree support, inline question answering, and Windows background server
The 0.154.0 stable release (npm-confirmed) adds experimental multi-worktree session support, the ability to answer clarification questions inline without losing the main draft, a shared background Codex server for Windows with daemon lifecycle commands, a Vim replace mode, and GPT-6-Astra availability in the model picker and Bedrock catalogs.
Source ↗ - releasenotice
Codex 0.153.4 fixes GPT-6-Astra model-picker visibility and makes it the default when no model is set
0.153.4 (2026-09-04) is the fourth same-day hotfix atop 0.153.0. It fixes Astra's absence from the bundled model picker (the defect underlying issue #42853 from last cycle) and changes behavior so that Astra becomes the bundled default model whenever no model is explicitly configured. It also further qualifies Astra's async-clarification-question guidance based on whether the tool is available in the session. npm confirms 0.153.4 as latest as of 2026-09-04T23:31Z.
Source ↗ - release
GPT-6-Astra added to Amazon Bedrock model catalog in 0.153.3
0.153.3 extends GPT-6-Astra availability to Amazon Bedrock Mantle and Runtime global/US routes, relevant to teams running Codex against Bedrock-hosted models on their own infrastructure.
Source ↗ - release
GPT-6-Astra made configurable via API in 0.153.1, ahead of model-picker exposure
The 0.153.1 hotfix lets sessions configure the new GPT-6-Astra model through the API/config without it appearing in the model picker or changing the default model. This is a narrower, API-first rollout path that BYO-key users should be aware of before the model is generally selectable.
Source ↗ - release
0.152.1 patch: Guardian honors Node REPL policies from model metadata
0.152.1, released the same day as 0.152.0, is a single-fix patch release for Guardian approval review handling of Node REPL policies sourced from model metadata.
Source ↗ - releasenotice
0.152.0 released: planning tool disabled by default, Vim search, credential-refresh progress
0.152.0 (2026-09-01) is the first release in this cycle with a real changelog. Notable for existing setups: the planning tool (tools.update_plan) is now disabled by default and must be re-enabled via config.toml. Also added: Vim-mode search in the composer, rate-limit banners with usage/credit actions, visible credential-refresh progress (including Amazon Bedrock reauth), MCP server names supporting ':'/'@'/'/'/'.', per-tool output_token_limit for MCP tools, and configurable thread/shellCommand timeouts beyond one hour. Bug fixes include preserving Guardian approval-review context across history compaction and fixing Windows sandbox execution with Microsoft Store PowerShell.
Source ↗ - release
Codex CLI 0.151.0 stable released
0.151.0 shipped 2026-08-29, three days after 0.150.1. Notable changes: configurable MCP-server discovery grace period, extensions can now inspect/replace MCP tool results before they reach the model, per-repository plugin catalog config, and nested subagent token usage now counts toward root goal budgets. Also fixes model/reasoning-effort correctness on fallback and several remote-sandbox path/OS enforcement issues. npm registry confirms 0.151.0 as latest, published 2026-08-29T09:59:26Z.
Source ↗ - release
Codex 0.150.1 bug fix — remote compaction now budgets retained images
Patch release 0.150.1 backports retained-image token budgeting to remote compaction so older images are trimmed to fit the budget.
Source ↗ - release
Codex 0.150.0 released with @-mentions for tasks, /copy picker, automatic task titles, Markdown link rendering, permission-mode shortcuts, and new Interrupt hooks
The 0.150.0 stable release (2026-08-26) ships several user-facing features: reference other Codex tasks with @-mentions, a /copy picker for full responses/code blocks/blockquotes, automatic descriptive titles for unnamed tasks with /rename suggestions, Markdown links rendered as clickable labels, keybindings to cycle permission modes plus '.' in Vim mode to repeat the last edit, and new 'Interrupt' hooks that fire on top-level turn interrupts. Bug fixes touch untrusted-project AGENTS.md handling, credential redaction, remote MCP bearer-token lookup, Windows sandbox paths under Unicode user dirs, Unix shutdown hangs from detached processes, and conversation compaction for Amazon Bedrock models.
Source ↗ - release
0.148.0-alpha.9
Release 0.148.0-alpha.9
Source ↗ - release
0.148.0-alpha.8
Release 0.148.0-alpha.8
Source ↗ - release
0.148.0-alpha.7
Release 0.148.0-alpha.7
Source ↗ - release
0.147.0-alpha.6.6
Release 0.147.0-alpha.6.6
Source ↗ - release
0.148.0-alpha.6
Release 0.148.0-alpha.6
Source ↗ - release
0.148.0-alpha.5
Release 0.148.0-alpha.5
Source ↗ - release
0.148.0-alpha.4
Release 0.148.0-alpha.4
Source ↗ - release
0.148.0-alpha.2
Release 0.148.0-alpha.2
Source ↗ - release
0.148.0-alpha.1
Release 0.148.0-alpha.1
Source ↗ - release
0.147.0
## New Features - Install portable Agent Plugins and search across local, personal, workspace, and remote plugin catalogs. (#36544, #36409, #36919, #36796) - Organize conversations into persistent, manually ordered sections and browse long transcripts incrementally. (#35722, #36007, #36380, #36948, #36950) - Enable automatically reviewed approvals with the new `--approve-for-me` CLI flag. (#36373) - Import Cursor-managed skills and synchronize changes to imported Claude and Cursor conversations without creating duplicates. (#36361, #36356, #35623) - Support the opt-in MCP 2026-07-28 protocol, including paginated discovery, multi-round requests, and non-blocking server startup. (#35724, #35725, #35590, #35742) - Enable cached web search and remote conversation compaction for Amazon Bedrock. (#36938, #36981) ## Bug Fixes - Redact secrets and complete bearer tokens from displayed commands and replayed conversation history. (#36893, #36908) - Prevent lost or stalled terminal input when focus returns, MCP servers initialize, or Ghostty handles keyboard shortcuts. (#35649, #35957, #36834) - Correct rendering and cursor positioning for Japanese characters, emoji, hyperlinks, and text near viewport boundaries. (#35960, #35962, #37166) - Properly interrupt Windows background processes and handle Windows filesystem paths consistently. (#35655, #35851, #37129) - Require explicit trust for unfamiliar local projects and enforce managed authentication restrictions before credentials are used. (#36960, #37132) - Harden plugin isolation and deny network access when policy updates fail. (#37027, #36967, #36037) ## Documentation - Improve the bundled OpenAI documentation skill with targeted official-source lookup and clearer guidance for Codex, model selection, and API workflows. (#36014) ## Chores - Upgrade the MCP SDK to 3.0.0, Ratatui to 0.30.2, and V8 to 150.4.0. (#36001, #35959, #35831) - Secure macOS release notarization using Azure Key Vault instead of exporting private sig
Source ↗ - release
0.147.0-alpha.13
Release 0.147.0-alpha.13
Source ↗ - release
0.147.0-alpha.12
Release 0.147.0-alpha.12
Source ↗ - release
0.147.0-alpha.6.5
Release 0.147.0-alpha.6.5
Source ↗ - release
0.147.0-alpha.11
Release 0.147.0-alpha.11
Source ↗ - release
0.146.1
## Bug Fixes - Apply safer automatic-review defaults for cyber-capable models and explain permission changes in the terminal interface. (#37057) ## Changelog Full Changelog: https://github.com/openai/codex/compare/rust-v0.146.0...rust-v0.146.1 - #37057 [0.146] Backport safer cyber-model auto-review defaults @anp-oai
Source ↗ - release
0.147.0-alpha.10
Release 0.147.0-alpha.10
Source ↗ - release
0.147.0-alpha.6.4
Release 0.147.0-alpha.6.4
Source ↗ - release
0.147.0-alpha.6.3
Release 0.147.0-alpha.6.3
Source ↗ - release
0.147.0-alpha.6.1
Release 0.147.0-alpha.6.1
Source ↗ - release
0.147.0-alpha.7
Release 0.147.0-alpha.7
Source ↗
Running Codex yourself
Compare it against the alternatives, or deploy it on a managed VPS in one click.