CodexLauncher v0.2.1

CodexLauncher v0.2.1 is a Windows provider-save stability release.

Highlights

• Fixes the Windows crash that could occur while saving a newly created model provider.
• Keeps an unsaved provider as a new draft and preserves the form when native validation or storage reports an error.
• Converts native provider-save panics into visible UI errors and prevents duplicate save operations while a request is running.
• Hardens Windows DPAPI output handling and adds Windows regression tests for token encryption and a complete local-token provider save.

Downloads

• macOS Apple Silicon: CodexLauncher-v0.2.1.dmg
• Windows x64 standard installer: CodexLauncher.Windows_0.2.1_x64-setup.exe
• Windows x64 MSI: CodexLauncher.Windows_0.2.1_x64_en-US.msi
• Windows x64 portable: CodexLauncher-Windows_0.2.1_x64-portable.zip

Signing notice

The macOS app uses an ad-hoc signature and is not notarized. The Windows packages are not Authenticode-signed. macOS Gatekeeper or Microsoft Defender SmartScreen may therefore show a warning on first launch.

CodexLauncher v0.2.0

CodexLauncher v0.2.0 is the first cross-platform release.

Highlights

• Adds the Windows Tauri app with CurrentUser DPAPI token storage, ChatGPT Store-app discovery, and NSIS, MSI, and portable ZIP packages.
• Uses model_messages.instructions_template on macOS and Windows while keeping the legacy base_instructions mirror for compatibility.
• Defaults custom catalog reasoning effort to medium and advertises minimal, low, medium, and high, avoiding unsupported reasoning.effort = "none" requests on OpenAI-compatible servers such as SGLang.
• Adds an editable, scrollable instructions template field and reset-to-official-template action to the macOS catalog editor.

Downloads

• macOS Apple Silicon: CodexLauncher-v0.2.0.dmg
• Windows x64 standard installer: CodexLauncher.Windows_0.2.0_x64-setup.exe
• Windows x64 MSI: CodexLauncher.Windows_0.2.0_x64_en-US.msi
• Windows x64 portable: CodexLauncher-Windows_0.2.0_x64-portable.zip

Signing notice

The macOS app uses an ad-hoc signature and is not notarized. The Windows packages are not Authenticode-signed. macOS Gatekeeper or Microsoft Defender SmartScreen may therefore show a warning on first launch.

CodexLauncher v0.1.8

What’s new

• Check GitHub Releases automatically at launch, limited to once every 24 hours.
• Add a manual 检查更新… command to the application menu.
• Download the release DMG, verify GitHub’s SHA-256 asset digest, and open the verified disk image.
• Compare semantic versions including v-prefixed, abbreviated, and prerelease versions.

Install

Open CodexLauncher-v0.1.8.dmg, then drag CodexLauncher.app to Applications.

This build is ad-hoc signed and is not notarized.

CodexLauncher v0.1.7

Changes

• Fetch provider model lists through a non-persistent, cache-disabled URL session.
• Prevent stale redirects and authenticated model-list responses from being reused from the on-disk URL cache.
• Purge legacy shared URL cache entries that may contain provider Authorization headers.
• Add regression coverage for ephemeral session settings and no-cache model requests.

Installation

Download the DMG and drag CodexLauncher into Applications, or extract the ZIP.

This build uses an ad-hoc signature and is not Apple-notarized because a Developer ID Application certificate is not currently available. On a new Mac, Gatekeeper may require right-clicking the app and choosing Open, or allowing it in Privacy & Security.

CodexLauncher v0.1.6

Changes

• Generate new local model catalog entries with direct-tool compatibility defaults.
• Add Model Catalog controls for tool mode, shell type, apply-patch type, multi-agent version, parallel tool calls, Tool Search, Responses Lite, modalities, web-search type, and truncation mode.
• Add a one-click compatibility reset using tool_mode = direct, apply_patch_tool_type = null, and use_responses_lite = false.
• Add regression tests for defaults, compatibility reset, and catalog round-tripping.

Installation

Download the DMG and drag CodexLauncher into Applications, or extract the ZIP.

This build uses an ad-hoc signature and is not Apple-notarized because a Developer ID Application certificate is not currently available. On a new Mac, Gatekeeper may require right-clicking the app and choosing Open, or allowing it in Privacy & Security.

CodexLauncher v0.1.5

• Fix third-party provider token delivery in the current ChatGPT/Codex runtime by using command-backed authentication.
• Keep static provider tokens in CodexLauncher private Application Support JSON without Keychain prompts.
• Automatically migrate locally stored tokens from env_key injection to the managed token helper.
• Separate Local Token, Environment, and Command authentication modes.
• Ship a drag-to-Applications DMG installer.

CodexLauncher v0.1.4

What's new

• Add command-backed bearer token authentication for custom model providers.
• Preserve unknown provider keys and nested provider tables when editing or renaming providers.
• Support provider query parameters, static headers, environment-backed headers, and command authentication when fetching /models.
• Replace Keychain token storage with a private provider-secrets.json file under Application Support to avoid repeated macOS password prompts.
• Keep auth.json reserved for OpenAI/ChatGPT login credentials.
• Show version 0.1.4 in the macOS About window and bundle metadata.

Token storage change

Existing legacy Keychain entries are not read, migrated, or deleted automatically. Re-enter each provider token once; it will be stored at ~/Library/Application Support/CodexLauncher/provider-secrets.json with file mode 600.

SHA-256: 17ee5a17bb9f52a68773c065f92ccdbd90a88d988654d55098972e541df23db9

CodexLauncher v0.1.3

• Launch the renamed /Applications/ChatGPT.app while retaining compatibility with the legacy /Applications/Codex.app.\n- Keep model provider display names independent from the selected model; provider names come only from each provider's own name or id.\n\nSHA-256: 0d9498c38ea70d826b729c82e5bbdd6bfad2ff84db6cd15e958102cc5d3c7c12

CodexLauncher v0.1.2

v0.1.2 - 2026-07-07

• Fix a crash when switching a profile between providers after editing provider settings.
• Make model catalog editor bindings resilient when the selected provider reloads a different model JSON.
• Generate a default provider env_key from the provider id when a Keychain token exists and env_key is blank.
• Clarify provider token and env_key behavior in the README.

Validation:

• swift build
• swift build -c release
• unzip -t dist/CodexLauncher-v0.1.2.zip

CodexLauncher v0.1.1

Changes

• Stop automatically renaming model providers when saving profiles or reloading config.
• Show provider display names in the profile model_provider picker, with the provider id kept as secondary context when different.
• Update release packaging docs for the v0.1.1 zip asset.