43 of 51 criteria have a sourced finding; 40 were also cross-checked by a second agent. Every comparison that includes OpenAI Codex links here for scope, limitations, atomic answers and sources.
Humans approve the rubric, templates and prototype; they do not review every finding. An agent source cross-check is not a hands-on product test. A missing sourced finding is not evidence that a capability is absent.
Codex is available inside the ChatGPT desktop app on macOS and Windows; Linux is a preview on the distributions documented separately. The shared application is named ChatGPT desktop app; its exact instruction is "Choose ChatGPT or Codex." Codex remains a separately selectable developer experience, so this is evidence of a shared desktop surface rather than an explicit product-wide rename or merger announcement. Codex remote access supports ChatGPT iOS and Android apps, subject to rollout; connected execution hosts must stay awake and online. Codex Cloud tasks are available through ChatGPT on the web. Codex CLI has macOS, Linux and Windows installers. Codex has an IDE extension with in-place code context and edits; compatibility details are bounded by each editor's installation guide. iOS Codex can start tasks, send instructions, approve requested actions and review diffs or test results. Remote clients receive notifications on completion or when input is needed. iOS can continue a cloud task or one on a connected Mac or Windows PC under the same account and workspace.
Scope
Surfaces: Codex CLI, Codex cloud, Codex IDE extension, Codex in ChatGPT desktop · Version: Public sources checked 2026-11-07; no installed release tested · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local, managed-cloud · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-07
Unknown after the targeted first-party documentation, release and repository pass: mobile_browser. No configured product fixture was run; documentation does not establish successful hands-on behavior.
ChatGPT desktop app is the shared shell name. Codex remains a product-selector choice; the checked docs do not establish a formal Codex product rename, merger or transition date. Generic ChatGPT Work capabilities are not inherited.
9 atomic answers
Desktop os [supported] Codex is available inside the ChatGPT desktop app on macOS and Windows; Linux is a preview on the distributions documented separately. The shared application is named ChatGPT desktop app; its exact instruction is "Choose ChatGPT or Codex." Codex remains a separately selectable developer experience, so this is evidence of a shared desktop surface rather than an explicit product-wide rename or merger announcement. Sources: codex:gap07-desktop, codex:gap07-desktop-identity, codex:gap07-desktop-selector.
Mobile apps [supported] Codex remote access supports ChatGPT iOS and Android apps, subject to rollout; connected execution hosts must stay awake and online. Sources: codex:gap07-remote.
Browser [supported] Codex Cloud tasks are available through ChatGPT on the web. Sources: codex:nov07-cloud.
Mobile browser [unknown] Not established for mobile browser in the named first-party source pass. Product access and scenario tests were not run.
Cli [supported] Codex CLI has macOS, Linux and Windows installers. Sources: codex:nov07-cli.
Editor extensions [supported] Codex has an IDE extension with in-place code context and edits; compatibility details are bounded by each editor's installation guide. Sources: codex:nov07-ide.
Task actions [supported] iOS Codex can start tasks, send instructions, approve requested actions and review diffs or test results. Sources: codex:gap07-mobile.
Notifications [supported] Remote clients receive notifications on completion or when input is needed. Sources: codex:gap07-remote.
Cross device continuity [supported] iOS can continue a cloud task or one on a connected Mac or Windows PC under the same account and workspace. Sources: codex:gap07-mobile.
Codex supplies its own coding harness. In the inspected CLI/core, agent delegation creates Codex threads through AgentControl; model-provider configuration and MCP tools do not replace the native harness with another coding CLI.
Scope
Surfaces: Codex CLI, Codex IDE extension, Codex in ChatGPT desktop, Codex cloud · Version: Codex rust-v0.167.0 source pinned at 086ccc833bc9cf01024ad2dd6af7dbd421b6bf9a and current first-party docs checked 2026-11-13. No installed release tested. · Plan: First-party documentation; client, plan and execution restrictions are named per atomic answer. Unlisted parity is unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
Not separately cross-checked
Only the named documented or inspected parts are established; other atomic questions retain their recorded gaps.
5 atomic answers
Harnesses [supported] The product provides the native Codex CLI coding harness; external coding harness support is not established. Sources: codex:gap07-cli-native.
Model providers [supported] Codex supports configured provider base URLs and credentials, with Responses protocol compatibility required. Sources: codex:gap07-config.
Custom adapters [partial] Custom model-provider configuration supplies base URL, API-key environment and headers; this does not establish an external coding-harness adapter. Sources: codex:gap07-config.
Session integration [supported] The native Codex CLI retains resumable chats and can submit Codex Cloud tasks. Sources: codex:gap07-cli-native.
Mixed harness operation [unsupported] The pinned CLI/core has no native alternate coding-harness dispatch route; multi-agent workers use Codex AgentControl threads. Sources: codex:gap-native-runtime.
Supported local Codex clients can sign in with ChatGPT. API-key sign-in uses API billing rather than included ChatGPT usage. API-key authentication works for CLI, SDK and IDE but excludes cloud features such as GitHub review and Slack integration. Ollama and LM Studio are built-in local model providers that do not require OpenAI sign-in. An experimental app-server API switches between ChatGPT workspaces of one managed login; switching among separate accounts is not documented. The local CLI can run with an API key, Bedrock or Vertex credentials, or a local Ollama/LM Studio model; codex cloud needs a ChatGPT sign-in.
Scope
Surfaces: Codex CLI, Codex cloud, Codex in ChatGPT desktop, Codex IDE extension · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: vendor_credits. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
7 atomic answers
Agent subscriptions [supported] Supported local Codex clients can sign in with ChatGPT. Sources: codex:nov07-auth.
Api accounts [supported] API-key sign-in uses API billing rather than included ChatGPT usage. Sources: codex:nov07-auth.
Vendor credits [unknown] Not established for vendor credits in the named first-party source pass. Product access and scenario tests were not run.
Local models [supported] Ollama and LM Studio are built-in local model providers that do not require OpenAI sign-in. Sources: codex:b10c-providers.
Account switching [partial] An experimental app-server API switches between ChatGPT workspaces of one managed login; switching among separate accounts is not documented. Sources: codex:b10c-as-accounts.
Vendor account requirement [partial] The local CLI can run with an API key, Bedrock or Vertex credentials, or a local Ollama/LM Studio model; codex cloud needs a ChatGPT sign-in. Sources: codex:b10c-providers, codex:b10c-cloud-cli.
Provider restrictions [supported] API-key authentication works for CLI, SDK and IDE but excludes cloud features such as GitHub review and Slack integration. Sources: codex:gap07-pricing.
Install the CLI, open a project directory, run codex and authenticate; elapsed setup time was not tested. Local non-Git projects are supported for scheduled work; worktree mode specifically requires Git. Worktree instructions require a Git repository and do not require a hosted remote. Choose GitHub repositories, let Codex prepare dependencies and tools, review setup and test results, then publish the environment. Configure an SSH alias, verify login, install/authenticate Codex remotely, then add the host and remote project in Settings Connections. Cloud tasks use repositories from a connected GitHub or GitLab account, and codex cloud exec targets an environment and branch; other hosts are not documented. codex cloud tells signed-out users to run codex login, and codex doctor checks installation, config, auth and network.
Scope
Surfaces: Codex CLI, Codex cloud · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local, managed-cloud · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
7 atomic answers
First task setup [supported] Install the CLI, open a project directory, run codex and authenticate; elapsed setup time was not tested. Sources: codex:nov07-cli.
No git [supported] Local non-Git projects are supported for scheduled work; worktree mode specifically requires Git. Sources: codex:gap07-worktrees.
Local git without remote [supported] Worktree instructions require a Git repository and do not require a hosted remote. Sources: codex:gap07-worktrees.
Hosted git [partial] Cloud tasks use repositories from a connected GitHub or GitLab account, and codex cloud exec targets an environment and branch; other hosts are not documented. Sources: codex:b10c-cloud-doc, codex:b10c-cloud-cli.
Prerequisite guidance [supported] codex cloud tells signed-out users to run codex login, and codex doctor checks installation, config, auth and network. Sources: codex:b10c-cloud-cli, codex:b10c-cli-main.
Managed cloud setup [supported] Choose GitHub repositories, let Codex prepare dependencies and tools, review setup and test results, then publish the environment. Sources: codex:nov07-cloud.
Byo vps onboarding [supported] Configure an SSH alias, verify login, install/authenticate Codex remotely, then add the host and remote project in Settings Connections. Sources: codex:gap07-remote.
codex exec runs non-interactive tasks in scripts and CI. The app-server serves JSON-RPC over stdio, Unix sockets or WebSocket, with generated TypeScript and JSON Schema; the commands are labelled experimental. TypeScript @openai/codex-sdk (Node.js 18+) and Python openai-codex. The SDKs start, run, continue and stream threads with structured output; codex cloud submits, checks status, lists, diffs and applies tasks. ChatGPT sign-in by browser OAuth or device code, API keys, and file or keyring credential storage; remote exec-server connections use bearer tokens. app-server, remote-control, cloud and exec-server are labelled experimental; exec and the SDKs carry no experimental label.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: mcp_server, webhooks. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
8 atomic answers
Cli [supported] codex exec runs non-interactive tasks in scripts and CI. Sources: codex:nov07-cli.
Api [partial] The app-server serves JSON-RPC over stdio, Unix sockets or WebSocket, with generated TypeScript and JSON Schema; the commands are labelled experimental. Sources: codex:b10c-appserver-main, codex:b10c-cli-main.
Mcp server [unknown] Not established for mcp server in the named first-party source pass. Product access and scenario tests were not run.
Webhooks [unknown] Not established for webhooks in the named first-party source pass. Product access and scenario tests were not run.
Operations [supported] The SDKs start, run, continue and stream threads with structured output; codex cloud submits, checks status, lists, diffs and applies tasks. Sources: codex:b10c-sdk-ts, codex:b10c-cloud-cli.
Authentication [supported] ChatGPT sign-in by browser OAuth or device code, API keys, and file or keyring credential storage; remote exec-server connections use bearer tokens. Sources: codex:b10c-login, codex:b10c-cred-store, codex:b10c-exec-server.
Stability [partial] app-server, remote-control, cloud and exec-server are labelled experimental; exec and the SDKs carry no experimental label. Sources: codex:b10c-cli-main, codex:b10c-daemon.
The CLI supports packaged skills and plugins. codex mcp manages external MCP servers, and /mcp login signs in from a session. SKILL.md skills (including project .agents/skills) and lifecycle hooks extend the agent alongside MCP tools. AGENTS.md and AGENTS.override.md are loaded, and a TUI migration flow imports configuration from other coding agents; importing their memories is still an off-by-default flag under development. MCP servers can use OAuth, including pre-registered client secrets; enterprise MCP can use Codex account grants. MCP config sets per-server and per-tool approval modes plus tool allow and deny lists; approval prompts include tool names and annotation hints. Users configure integrations in config.toml; admins can restrict hooks to managed ones via requirements.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
7 atomic answers
Mcp client [supported] codex mcp manages external MCP servers, and /mcp login signs in from a session. Sources: codex:b10c-mcp-cmd, codex:b10c-rel-0161.
Plugins [supported] The CLI supports packaged skills and plugins. Sources: codex:nov07-cli.
Custom tools [supported] SKILL.md skills (including project .agents/skills) and lifecycle hooks extend the agent alongside MCP tools. Sources: codex:b10c-skills.
Instructions import [supported] AGENTS.md and AGENTS.override.md are loaded, and a TUI migration flow imports configuration from other coding agents; importing their memories is still an off-by-default flag under development. Sources: codex:b10c-agents-md, codex:b10c-ext-migration, codex:b10c-features.
Permission scope [supported] MCP config sets per-server and per-tool approval modes plus tool allow and deny lists; approval prompts include tool names and annotation hints. Sources: codex:b10c-mcp-tools, codex:b10c-rel-0167.
Integration owner [partial] Users configure integrations in config.toml; admins can restrict hooks to managed ones via requirements. Sources: codex:b10c-config-md.
Client implementation: The desktop and mobile apps act as remote clients of the app-server daemon; their own UI technology is not in the public source. Pinned source ccf9be8 holds the Rust CLI/TUI and only a desktop launcher; desktop and IDE client code is not public. Not established: desktop implementation, mobile implementation.
Scope
Surfaces: Codex CLI, Codex in ChatGPT desktop · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: desktop_implementation, mobile_implementation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
4 atomic answers
Desktop implementation [unknown] Not established for desktop implementation in the named first-party source pass. Product access and scenario tests were not run.
Mobile implementation [unknown] Not established for mobile implementation in the named first-party source pass. Product access and scenario tests were not run.
Hybrids [partial] The desktop and mobile apps act as remote clients of the app-server daemon; their own UI technology is not in the public source. Sources: codex:b10c-daemon.
Implementation evidence [partial] Pinned source ccf9be8 holds the Rust CLI/TUI and only a desktop launcher; desktop and IDE client code is not public. Sources: codex:b10c-desktop.
Intake and external tracker relationship: Tasks are threads started from the CLI, SDK or codex cloud exec; there is no separate issue object. Prompts come as arguments or stdin, with optional images. Sessions have IDs or names, and cloud tasks have task IDs. Not established: external import, chat git triggers, authoritative tracker, sync, supported edits, external roundtrips.
Scope
Surfaces: Codex CLI, Codex cloud · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: external_import, chat_git_triggers, authoritative_tracker, sync, supported_edits, external_roundtrips. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
9 atomic answers
Local creation [partial] Tasks are threads started from the CLI, SDK or codex cloud exec; there is no separate issue object. Sources: codex:b10c-exec-cli, codex:b10c-cloud-cli.
Prompt intake [supported] Prompts come as arguments or stdin, with optional images. Sources: codex:b10c-exec-cli.
External import [unknown] Not established for external import in the named first-party source pass. Product access and scenario tests were not run.
Chat git triggers [unknown] Not established for chat git triggers in the named first-party source pass. Product access and scenario tests were not run.
Authoritative tracker [unknown] Not established for authoritative tracker in the named first-party source pass. Product access and scenario tests were not run.
Sync [unknown] Not established for sync in the named first-party source pass. Product access and scenario tests were not run.
Supported edits [unknown] Not established for supported edits in the named first-party source pass. Product access and scenario tests were not run.
External roundtrips [unknown] Not established for external roundtrips in the named first-party source pass. Product access and scenario tests were not run.
Durable task plans: The update_plan tool publishes and updates a step plan in the thread. Thread goals persist in the local state database and saved sessions resume; step plans live in the session transcript. Not established: acceptance criteria, editors, revisions, implementation links, amendment.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: acceptance_criteria, editors, revisions, implementation_links, amendment. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
7 atomic answers
Plans [partial] The update_plan tool publishes and updates a step plan in the thread. Sources: codex:b10c-plan.
Acceptance criteria [unknown] Not established for acceptance criteria in the named first-party source pass. Product access and scenario tests were not run.
Editors [unknown] Not established for editors in the named first-party source pass. Product access and scenario tests were not run.
Revisions [unknown] Not established for revisions in the named first-party source pass. Product access and scenario tests were not run.
Implementation links [unknown] Not established for implementation links in the named first-party source pass. Product access and scenario tests were not run.
Amendment [unknown] Not established for amendment in the named first-party source pass. Product access and scenario tests were not run.
Restart persistence [partial] Thread goals persist in the local state database and saved sessions resume; step plans live in the session transcript. Sources: codex:b10c-goals, codex:b10c-rollout.
Hierarchy, dependencies and evolving scope: Subagent threads persist under parents and come back when a parent is unarchived; this is agent nesting, not a task hierarchy. Not established: epics, blockers, readiness, enforcement, state propagation, task to epic, identity retention, workflow constraints.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: epics, blockers, readiness, enforcement, state_propagation, task_to_epic, identity_retention, workflow_constraints. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
9 atomic answers
Epics [unknown] Not established for epics in the named first-party source pass. Product access and scenario tests were not run.
Hierarchy [partial] Subagent threads persist under parents and come back when a parent is unarchived; this is agent nesting, not a task hierarchy. Sources: codex:b10c-multi-agents, codex:b10c-rel-0166.
Blockers [unknown] Not established for blockers in the named first-party source pass. Product access and scenario tests were not run.
Readiness [unknown] Not established for readiness in the named first-party source pass. Product access and scenario tests were not run.
Enforcement [unknown] Not established for enforcement in the named first-party source pass. Product access and scenario tests were not run.
State propagation [unknown] Not established for state propagation in the named first-party source pass. Product access and scenario tests were not run.
Task to epic [unknown] Not established for task to epic in the named first-party source pass. Product access and scenario tests were not run.
Identity retention [unknown] Not established for identity retention in the named first-party source pass. Product access and scenario tests were not run.
Workflow constraints [unknown] Not established for workflow constraints in the named first-party source pass. Product access and scenario tests were not run.
Shared context and memory: Codex reads AGENTS.md before beginning work. SKILL.md skills, including project .agents/skills, plus bundled system skills. AGENTS.md files in the repository act as project docs; no other document store was found. A two-phase memory pipeline is stable but off by default in the feature table; when memories are enabled, fresh profiles default to memory v2. Global guidance is followed by repository-root to working-directory instructions; nearer files override earlier guidance. The memory read path parses memory citations; what users see was not established. Edit or remove an AGENTS.override.md file before the next startup to change guidance. Instruction discovery runs once per session startup; automatic rereading during a run is not established. Not established: deletion, takeover continuity.
Scope
Surfaces: Codex CLI, Codex IDE extension, Codex in ChatGPT desktop, Codex cloud · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: First-party documentation; client, plan and execution restrictions are named per atomic answer. Unlisted parity is unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: deletion, takeover_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
10 atomic answers
Instructions [supported] Codex reads AGENTS.md before beginning work. Sources: codex:gap07-agents-md.
Skills [supported] SKILL.md skills, including project .agents/skills, plus bundled system skills. Sources: codex:b10c-skills.
Project docs [partial] AGENTS.md files in the repository act as project docs; no other document store was found. Sources: codex:b10c-agents-md.
Scope [supported] Global guidance is followed by repository-root to working-directory instructions; nearer files override earlier guidance. Sources: codex:gap07-agents-md.
Provenance [partial] The memory read path parses memory citations; what users see was not established. Sources: codex:b10c-memories.
Correction [supported] Edit or remove an AGENTS.override.md file before the next startup to change guidance. Sources: codex:gap07-agents-md.
Deletion [unknown] Not established for deletion in the named first-party source pass. Product access and scenario tests were not run.
Freshness [partial] Instruction discovery runs once per session startup; automatic rereading during a run is not established. Sources: codex:gap07-agents-md.
Takeover continuity [unknown] Not established for takeover continuity in the named first-party source pass. Product access and scenario tests were not run.
codex --search enables live web search with visible search activity. codex --image accepts images with a prompt. A file-search crate and shell tools search the local workspace.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: indexing, freshness, access_scope, api_retrieval, citations, conflict_detection, review_enforcement. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
Indexing [unknown] Not established for indexing in the named first-party source pass. Product access and scenario tests were not run.
External knowledge [supported] codex --search enables live web search with visible search activity. Sources: codex:nov07-cli.
Multimodal inputs [supported] codex --image accepts images with a prompt. Sources: codex:nov07-cli.
Freshness [unknown] Not established for freshness in the named first-party source pass. Product access and scenario tests were not run.
Access scope [unknown] Not established for access scope in the named first-party source pass. Product access and scenario tests were not run.
Api retrieval [unknown] Not established for api retrieval in the named first-party source pass. Product access and scenario tests were not run.
Citations [unknown] Not established for citations in the named first-party source pass. Product access and scenario tests were not run.
Conflict detection [unknown] Not established for conflict detection in the named first-party source pass. Product access and scenario tests were not run.
Review enforcement [unknown] Not established for review enforcement in the named first-party source pass. Product access and scenario tests were not run.
The public CLI supplies a per-turn update_plan checklist inside a conversation event stream. It does not establish separate persistent user-managed issues/tasks across sessions. The hosted Cloud and closed client implementation remains unestablished.
Scope
Surfaces: Codex CLI · Version: Codex rust-v0.167.0 source pinned at 086ccc833bc9cf01024ad2dd6af7dbd421b6bf9a and current first-party docs checked 2026-11-13. No installed release tested. · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-13 (9 days old)
Agent cross-check
Not separately cross-checked
Only the named documented or inspected parts are established; other atomic questions retain their recorded gaps.
17 atomic answers
Create [unknown] Not established for create in the named first-party source pass. Product access and scenario tests were not run.
Edit [unknown] Not established for edit in the named first-party source pass. Product access and scenario tests were not run.
Discuss [unknown] Not established for discuss in the named first-party source pass. Product access and scenario tests were not run.
Prioritize [unknown] Not established for prioritize in the named first-party source pass. Product access and scenario tests were not run.
Label [unknown] Not established for label in the named first-party source pass. Product access and scenario tests were not run.
Assign [unknown] Not established for assign in the named first-party source pass. Product access and scenario tests were not run.
Search [unknown] Not established for search in the named first-party source pass. Product access and scenario tests were not run.
Filter [unknown] Not established for filter in the named first-party source pass. Product access and scenario tests were not run.
Archive [unknown] Not established for archive in the named first-party source pass. Product access and scenario tests were not run.
Reopen [unknown] Not established for reopen in the named first-party source pass. Product access and scenario tests were not run.
Project epic views [unknown] Not established for project epic views in the named first-party source pass. Product access and scenario tests were not run.
Idea parking [unknown] Not established for idea parking in the named first-party source pass. Product access and scenario tests were not run.
Explicit dispatch [unknown] Not established for explicit dispatch in the named first-party source pass. Product access and scenario tests were not run.
Task session separation [unsupported] The update_plan items are conversation/turn plan events, not independent issue records. Sources: codex:gap-plan-scope.
Metadata [unknown] Not established for metadata in the named first-party source pass. Product access and scenario tests were not run.
Durable states [partial] Plan steps have pending, in-progress and completed states in the thread event stream; independent work-item state is not established. Sources: codex:gap-plan-scope.
External tracker requirement [unknown] Not established for external tracker requirement in the named first-party source pass. Product access and scenario tests were not run.
Each cloud task has its own workspace; the page gives no concurrency quota. The agent command center browses and switches saved sessions, and up to 6 subagent threads run by default; concurrent execution of several top-level sessions was not established. Subagents default to 6 threads at depth 1, both configurable. Plus and Standard Business usage is estimated per five-hour period; Pro currently has no five-hour limit, weekly limits may apply, and Enterprise/Edu with flexible pricing have no fixed rate limits. No fixed concurrency quota is published. The command center marks unread results, list_agents reports live agents, and codex cloud status shows task state.
Scope
Surfaces: Codex cloud, Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: managed-cloud · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: observed_concurrency. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
5 atomic answers
Concurrent sessions [partial] The agent command center browses and switches saved sessions, and up to 6 subagent threads run by default; concurrent execution of several top-level sessions was not established. Sources: codex:b10c-cli-main, codex:b10c-agent-config.
Concurrent tasks [supported] Each cloud task has its own workspace; the page gives no concurrency quota. Sources: codex:nov07-cloud.
Quotas [partial] Subagents default to 6 threads at depth 1, both configurable. Plus and Standard Business usage is estimated per five-hour period; Pro currently has no five-hour limit, weekly limits may apply, and Enterprise/Edu with flexible pricing have no fixed rate limits. No fixed concurrency quota is published. Sources: codex:b10c-agent-config, codex:b10c-pricing, codex:c165-usage-limits.
Observed concurrency [unknown] Not established for observed concurrency in the named first-party source pass. Product access and scenario tests were not run.
Ownership and dispatch: spawn_agent takes an optional agent role; there is no human assignment or task claim object. User-defined roles carry descriptions that guide which role the model spawns, plus their own config layers; routing is model-driven. Not established: human assignment, claims, queues, duplicate prevention, ownership enforcement.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: human_assignment, claims, queues, duplicate_prevention, ownership_enforcement. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
7 atomic answers
Human assignment [unknown] Not established for human assignment in the named first-party source pass. Product access and scenario tests were not run.
Agent assignment [partial] spawn_agent takes an optional agent role; there is no human assignment or task claim object. Sources: codex:b10c-multi-agents, codex:b10c-roles.
Claims [unknown] Not established for claims in the named first-party source pass. Product access and scenario tests were not run.
Queues [unknown] Not established for queues in the named first-party source pass. Product access and scenario tests were not run.
Routing [partial] User-defined roles carry descriptions that guide which role the model spawns, plus their own config layers; routing is model-driven. Sources: codex:b10c-roles.
Duplicate prevention [unknown] Not established for duplicate prevention in the named first-party source pass. Product access and scenario tests were not run.
Ownership enforcement [unknown] Not established for ownership enforcement in the named first-party source pass. Product access and scenario tests were not run.
Local custom agents can set model and reasoning effort; unspecified children inherit their parent. The main agent collects subagent results into a consolidated response. Further scoped facts are recorded in the atomic answers. User-defined roles with descriptions, config layers and nicknames. send_input, send_message and followup_task deliver messages to existing agents. The parent thread orchestrates through tools; spawn depth defaults to 1. Subagent relationships persist, and unarchiving a parent restores descendant threads. followup_task and resume_agent continue existing agents. The model calls spawn_agent, usually when asked; the public source shows no direct user control to spawn a child.
Scope
Surfaces: Codex CLI, Codex in ChatGPT desktop, Codex IDE extension · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: cross_harness, authentication. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
12 atomic answers
Roles [supported] User-defined roles with descriptions, config layers and nicknames. Sources: codex:b10c-roles.
Model harness effort [supported] Local custom agents can set model and reasoning effort; unspecified children inherit their parent. Sources: codex:nov07-subagents.
Messages [supported] send_input, send_message and followup_task deliver messages to existing agents. Sources: codex:b10c-multi-agents.
Results return [supported] The main agent collects subagent results into a consolidated response. Sources: codex:nov07-subagents.
Cross harness [unknown] Not established for cross harness in the named first-party source pass. Product access and scenario tests were not run.
Activation [supported] Current local clients delegate on explicit request or applicable project or skill instructions; the docs say current releases enable subagents by default. Sources: codex:nov07-subagents.
Authentication [unknown] Not established for authentication in the named first-party source pass. Product access and scenario tests were not run.
Live progress [supported] CLI /subagents switches among child threads; desktop and eligible IDE UI expose child activity. Sources: codex:nov07-subagents.
Retained results [partial] Subagent relationships persist, and unarchiving a parent restores descendant threads. Sources: codex:b10c-rel-0167, codex:b10c-rel-0166.
Followup [supported] followup_task and resume_agent continue existing agents. Sources: codex:b10c-multi-agents.
Invocation [partial] The model calls spawn_agent, usually when asked; the public source shows no direct user control to spawn a child. Sources: codex:b10c-multi-agents.
Desktop Codex supports recurring tasks with an RRULE custom cadence. Standalone scheduled tasks start a new chat per run; an in-chat schedule retains the conversation. Local project schedules require the desktop app running on the execution computer. Local scheduled tasks require the computer on and app running; cloud web scheduling is a separate workflow. Thread goals hold an objective, an optional token budget and a status: active, paused, blocked (shown as stalled in the TUI), usage-limited (set when a usage limit stops the active goal), budget-limited (token budget reached) or complete.
Scope
Surfaces: Codex CLI, Codex IDE extension, Codex in ChatGPT desktop, Codex cloud · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: First-party documentation; client, plan and execution restrictions are named per atomic answer. Unlisted parity is unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: events, overlaps, retries. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
8 atomic answers
Schedules [supported] Desktop Codex supports recurring tasks with an RRULE custom cadence. Sources: codex:gap07-schedules.
Events [unknown] Not established for events in the named first-party source pass. Product access and scenario tests were not run.
Recurring tasks [supported] Standalone scheduled tasks start a new chat per run; an in-chat schedule retains the conversation. Sources: codex:gap07-schedules.
Completion conditions [partial] Thread goals hold an objective, an optional token budget and a status: active, paused, blocked (shown as stalled in the TUI), usage-limited (set when a usage limit stops the active goal), budget-limited (token budget reached) or complete. Sources: codex:b10c-goals, codex:b10c-features.
Scheduler location [supported] Local project schedules require the desktop app running on the execution computer. Sources: codex:gap07-schedules.
Overlaps [unknown] Not established for overlaps in the named first-party source pass. Product access and scenario tests were not run.
Retries [unknown] Not established for retries in the named first-party source pass. Product access and scenario tests were not run.
Client independence [partial] Local scheduled tasks require the computer on and app running; cloud web scheduling is a separate workflow. Sources: codex:gap07-schedules.
The interactive CLI accepts steering during the active turn. Scheduled acts as an inbox with unread indicators for findings requiring attention. Interrupting a turn also stops Code Mode cells and nested tool calls; agents can be interrupted or closed. codex resume, fork and queue --resume reopen sessions, and resume_agent reopens subagents. The TUI sends terminal desktop notifications (on by default), a notify command can be configured, and the command center marks unread results; push or email notifications are not in the public source. On-request approval is the default. The stable guardian_approval feature flag is on by default, which makes automatic review available, but approval requests go to the user unless approvals_reviewer is set to auto_review; managed checkpoints force a human decision. Safety blocks explain themselves and offer safer follow-ups; subagents receive their preparation failures. Sessions are saved under CODEX_HOME/sessions and can be archived, unarchived or deleted.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: pause, chat_terminal_parity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
10 atomic answers
Pause [unknown] Not established for pause in the named first-party source pass. Product access and scenario tests were not run.
Stop [supported] Interrupting a turn also stops Code Mode cells and nested tool calls; agents can be interrupted or closed. Sources: codex:b10c-rel-0163, codex:b10c-multi-agents.
Steer [supported] The interactive CLI accepts steering during the active turn. Sources: codex:nov07-cli.
Notifications [partial] The TUI sends terminal desktop notifications (on by default), a notify command can be configured, and the command center marks unread results; push or email notifications are not in the public source. Sources: codex:b10c-notify, codex:b10c-rel-0167.
Pending decisions [supported] Scheduled acts as an inbox with unread indicators for findings requiring attention. Sources: codex:gap07-schedules.
Escalation [supported] On-request approval is the default. The stable guardian_approval feature flag is on by default, which makes automatic review available, but approval requests go to the user unless approvals_reviewer is set to auto_review; managed checkpoints force a human decision. Sources: codex:b10c-approval, codex:b10c-features, codex:c165-approvals-reviewer, codex:b10c-as-threads.
Visible errors [partial] Safety blocks explain themselves and offer safer follow-ups; subagents receive their preparation failures. Sources: codex:b10c-rel-0166, codex:b10c-rel-0160.
Chat terminal parity [unknown] Not established for chat terminal parity in the named first-party source pass. Product access and scenario tests were not run.
Retained history [supported] Sessions are saved under CODEX_HOME/sessions and can be archived, unarchived or deleted. Sources: codex:b10c-rollout, codex:b10c-cli-main.
Fleet and epic management UI: Subagent threads are tied to parents and restored with them; the default TUI workspace shows subagent transcripts beside the main conversation. A nested tree view is not documented. The command center and session picker mark unread results. The command center pages through and searches older tasks, and tasks can be pinned. Directory grouping and a repository filter span sessions across linked worktrees. Not established: fleet fixture, dependencies, navigation errors, fixture mode.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: fleet_fixture, dependencies, navigation_errors, fixture_mode. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
8 atomic answers
Fleet fixture [unknown] Not established for fleet fixture in the named first-party source pass. Product access and scenario tests were not run.
Parent child [partial] Subagent threads are tied to parents and restored with them; the default TUI workspace shows subagent transcripts beside the main conversation. A nested tree view is not documented. Sources: codex:b10c-rel-0166, codex:b10c-rel-0171.
Dependencies [unknown] Not established for dependencies in the named first-party source pass. Product access and scenario tests were not run.
Attention [partial] The command center and session picker mark unread results. Sources: codex:b10c-rel-0167.
Workload [partial] Directory grouping and a repository filter span sessions across linked worktrees. Sources: codex:b10c-rel-0165.
Navigation errors [unknown] Not established for navigation errors in the named first-party source pass. Product access and scenario tests were not run.
Fixture mode [unknown] Not established for fixture mode in the named first-party source pass. Product access and scenario tests were not run.
A scheduled task can choose default or explicit model and reasoning effort. Desktop tasks select projects and local or worktree execution. A single codex exec prompt or one interactive turn. The model spawns subagents through tools when asked in chat. Reasoning effort can change mid-conversation by default, and follow-ups steer a running turn. Subagent relationships are persisted. /config searches settings and new sessions show occasional tips.
Scope
Surfaces: Codex CLI, Codex in ChatGPT desktop, Codex IDE extension · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: growing_epic, live_roles, auth_friction. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
10 atomic answers
Quick task [supported] A single codex exec prompt or one interactive turn. Sources: codex:b10c-exec-cli.
Growing epic [unknown] Not established for growing epic in the named first-party source pass. Product access and scenario tests were not run.
Presets [supported] A scheduled task can choose default or explicit model and reasoning effort. Sources: codex:gap07-schedules.
Natural language [partial] The model spawns subagents through tools when asked in chat. Sources: codex:b10c-multi-agents.
Agent configuration [supported] Desktop tasks select projects and local or worktree execution. Sources: codex:gap07-schedules.
In run changes [supported] Reasoning effort can change mid-conversation by default, and follow-ups steer a running turn. Sources: codex:b10c-rel-0165, codex:b10c-rel-0163.
Live roles [unknown] Not established for live roles in the named first-party source pass. Product access and scenario tests were not run.
Retained relationships [partial] Subagent relationships are persisted. Sources: codex:b10c-rel-0167.
The CLI inspects, edits and runs tools in the local project. Codex can execute against an SSH host with Codex installed and authenticated on its PATH. Cloud tasks can keep working while the originating computer is asleep. The desktop UI can control Codex tools and repository files on an SSH host. A paired iOS or Android app can steer and approve work on an awake Mac/Windows host. The Apache-2.0 agent runs on any machine the user controls; no self-hosted cloud service was found. Local use bills the signed-in ChatGPT plan or the API or cloud-provider account; cloud tasks need ChatGPT sign-in and share the plan allowance. A background server starts automatically for eligible interactive sessions and serves remote desktop and mobile clients; queued messages resume after reconnection. Continued execution after the client disconnects was not established.
Scope
Surfaces: Codex CLI, Codex cloud · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local, managed-cloud · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: provisioning, resources, region, expiry. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
12 atomic answers
Local [supported] The CLI inspects, edits and runs tools in the local project. Sources: codex:nov07-cli.
Byo remote [supported] Codex can execute against an SSH host with Codex installed and authenticated on its PATH. Sources: codex:gap07-remote.
Self hosted [partial] The Apache-2.0 agent runs on any machine the user controls; no self-hosted cloud service was found. Sources: codex:b10c-license, codex:b10c-providers.
Managed cloud [supported] Cloud tasks can keep working while the originating computer is asleep. Sources: codex:nov07-cloud.
Hybrid [supported] The desktop UI can control Codex tools and repository files on an SSH host. Sources: codex:gap07-remote.
Provisioning [unknown] Not established for provisioning in the named first-party source pass. Product access and scenario tests were not run.
Payer [partial] Local use bills the signed-in ChatGPT plan or the API or cloud-provider account; cloud tasks need ChatGPT sign-in and share the plan allowance. Sources: codex:b10c-providers, codex:b10c-cloud-cli, codex:b10c-pricing.
Resources [unknown] Not established for resources in the named first-party source pass. Product access and scenario tests were not run.
Region [unknown] Not established for region in the named first-party source pass. Product access and scenario tests were not run.
Expiry [unknown] Not established for expiry in the named first-party source pass. Product access and scenario tests were not run.
Client disconnect [partial] A background server starts automatically for eligible interactive sessions and serves remote desktop and mobile clients; queued messages resume after reconnection. Continued execution after the client disconnects was not established. Sources: codex:b10c-rel-0157, codex:b10c-rel-0160, codex:b10c-daemon.
Other device continuation [supported] A paired iOS or Android app can steer and approve work on an awake Mac/Windows host. Sources: codex:gap07-remote.
Codex supports local OS sandboxing, user-managed devcontainers and Git worktrees. The current Cloud environment guide explicitly gives a VM per cloud task with plan-dependent resources; local and cloud isolation scopes are separate. On Linux, bubblewrap adds a PID namespace and seccomp filters syscalls; macOS uses Seatbelt. Isolation between tasks was not established. Credentials are masked in supported remote shells and can live in the keyring. In Codex Cloud (Legacy) environments, which serve Code Review and the Linear and GitHub integrations, secrets are available only to setup scripts and are removed before the agent phase; secret handling in the current Cloud environments guide was not checked.
Scope
Surfaces: Codex CLI, Codex cloud, Codex IDE extension · Version: Earlier findings: Codex rust-v0.167.0 source pinned at 086ccc833bc9cf01024ad2dd6af7dbd421b6bf9a and current first-party docs checked 2026-11-13. No installed release tested.; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local, managed-cloud · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Only the named documented or inspected parts are established; other atomic questions retain their recorded gaps.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
12 atomic answers
Checkout [supported] Each cloud task has a separate workspace; hostile-code isolation was not tested. Sources: codex:nov07-cloud.
Branch [supported] A worktree can use a separate branch; the starting branch is chosen in the new-chat flow. Sources: codex:gap07-worktrees.
Worktree [supported] Codex supports a separate Git worktree for an independent chat. Sources: codex:gap07-worktrees.
Container [supported] The security guide documents running Codex inside a devcontainer, with optional inner Linux sandbox. Sources: codex:gap07-otel.
Vm [supported] Each Codex Cloud task runs in a VM. The guide lists CPU/RAM/disk allocations by ChatGPT plan. Sources: codex:gap-cloud-vm.
Os sandbox [supported] CLI and IDE use OS mechanisms to enforce sandbox policies; the documentation describes workspace-limited writes and network restrictions. Sources: codex:nov07-permissions.
Filesystem [supported] The CLI sandbox defaults to read-only; workspace-write allows writes only to the working directory and configured writable roots; danger-full-access removes the limit. Source-inspected at rust-v0.167.0; not tested. Sources: codex:src2-codex-sandbox-mode, codex:src2-codex-network.
Process [partial] On Linux, bubblewrap adds a PID namespace and seccomp filters syscalls; macOS uses Seatbelt. Isolation between tasks was not established. Sources: codex:b10c-sandbox.
Network [supported] In the workspace-write policy outbound network access is off by default and turned on with network_access. Source-inspected at rust-v0.167.0; not tested. Sources: codex:src2-codex-network.
Secrets [partial] Credentials are masked in supported remote shells and can live in the keyring. In Codex Cloud (Legacy) environments, which serve Code Review and the Linear and GitHub integrations, secrets are available only to setup scripts and are removed before the agent phase; secret handling in the current Cloud environments guide was not checked. Sources: codex:b10c-rel-0164, codex:b10c-cred-store, codex:b10c-cloud-env.
Code separation [supported] Worktrees have independent files while sharing Git metadata. Sources: codex:gap07-worktrees.
Security boundary [partial] A devcontainer protects the host but cannot protect credentials already available inside it when sandbox and approvals are bypassed. Sources: codex:gap07-otel.
Phone worktree control still requires the execution computer awake; this is distinct from cloud tasks that run while it sleeps. codex resume reopens saved local chats; this does not establish worker-crash recovery. Remote control stops if the host sleeps, disconnects or closes the app and becomes available when the host returns; successful recovery was not tested. Handoff moves a chat between its Local checkout and a worktree while Codex performs the Git operations. An existing cloud task resumes with its own saved files, uncommitted changes and installed tools. /daemon restarts the background server, warns about interruptible work and reconnects. Context is compacted automatically, and a failed compaction leaves no partial summaries. Uncertain submissions are resolved before queued messages resend. codex doctor diagnoses problems, and damaged SQLite databases are kept as backups.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: worker_crash, expiry, multi_day_continuity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
12 atomic answers
Client disconnect [partial] Phone worktree control still requires the execution computer awake; this is distinct from cloud tasks that run while it sleeps. Sources: codex:gap07-worktrees.
Worker crash [unknown] Not established for worker crash in the named first-party source pass. Product access and scenario tests were not run.
Service restart [partial] /daemon restarts the background server, warns about interruptible work and reconnects. Sources: codex:b10c-rel-0163.
Expiry [unknown] Not established for expiry in the named first-party source pass. Product access and scenario tests were not run.
Transcript reload [supported] codex resume reopens saved local chats; this does not establish worker-crash recovery. Sources: codex:nov07-cli.
Process reconnect [partial] Remote control stops if the host sleeps, disconnects or closes the app and becomes available when the host returns; successful recovery was not tested. Sources: codex:gap07-remote.
Task resume [supported] Handoff moves a chat between its Local checkout and a worktree while Codex performs the Git operations. Sources: codex:gap07-worktrees.
State reconstruction [supported] An existing cloud task resumes with its own saved files, uncommitted changes and installed tools. Sources: codex:gap07-environments.
Context loss [partial] Context is compacted automatically, and a failed compaction leaves no partial summaries. Sources: codex:b10c-rel-0164.
Duplicate actions [partial] Uncertain submissions are resolved before queued messages resend. Sources: codex:b10c-rel-0160.
Manual repair [partial] codex doctor diagnoses problems, and damaged SQLite databases are kept as backups. Sources: codex:b10c-cli-main, codex:b10c-rel-0161.
Multi day continuity [unknown] Not established for multi day continuity in the named first-party source pass. Product access and scenario tests were not run.
Cloud setup identifies and installs repository runtimes, dependencies and tools. Background repository refresh preserves dependency caches without rerunning installation. Publish captures prepared filesystem state for new tasks. The start skill documents how to start services and check readiness. New tasks reuse the published prepared filesystem; background refresh preserves caches. Transient environment startup failures retry on the next turn.
Scope
Surfaces: Codex cloud, Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: managed-cloud · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: ports, secrets, collisions, remote_setup. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
10 atomic answers
Dependencies [supported] Cloud setup identifies and installs repository runtimes, dependencies and tools. Sources: codex:gap07-environments.
Provider SSE idle timeout defaults to 300,000 ms and is configurable. Provider HTTP request retry count defaults to four; interrupted SSE streams default to five retries. Retries for HTTP and SSE failures are configurable; successful recovery was not tested. Automatic compaction, a context-remaining tool and optional goal token budgets; no hard per-task cap was found.
Scope
Surfaces: Codex CLI, Codex IDE extension, Codex in ChatGPT desktop, Codex cloud · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: First-party documentation; client, plan and execution restrictions are named per atomic answer. Unlisted parity is unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: time_limits, turn_limits, spend_limits, enforcement. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
8 atomic answers
Time limits [unknown] Not established for time limits in the named first-party source pass. Product access and scenario tests were not run.
Turn limits [unknown] Not established for turn limits in the named first-party source pass. Product access and scenario tests were not run.
Opt-in OTel exports structured API, tool, approval and stream events. Telemetry records request and tool durations and status; full distributed-trace parity is not implied. API events include status, duration and error details. Configured OTel stream events include token counts at response completion. Tool-decision events identify approval or denial and whether configuration or user made the decision.
Scope
Surfaces: Codex CLI, Codex IDE extension, Codex in ChatGPT desktop, Codex cloud · Version: Public sources checked 2026-11-07; no installed release tested · Plan: First-party documentation; client, plan and execution restrictions are named per atomic answer. Unlisted parity is unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-07
Unknown after the targeted first-party documentation, release and repository pass: task_stats, agent_stats, project_stats, time_stats, failure_status, reconnect_consistency, performance_method, responsiveness. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Traces [partial] Telemetry records request and tool durations and status; full distributed-trace parity is not implied. Sources: codex:gap07-otel.
Failure reasons [supported] API events include status, duration and error details. Sources: codex:gap07-otel.
Usage export [supported] Configured OTel stream events include token counts at response completion. Sources: codex:gap07-otel.
Task stats [unknown] Not established for task stats in the named first-party source pass. Product access and scenario tests were not run.
Agent stats [unknown] Not established for agent stats in the named first-party source pass. Product access and scenario tests were not run.
Project stats [unknown] Not established for project stats in the named first-party source pass. Product access and scenario tests were not run.
Time stats [unknown] Not established for time stats in the named first-party source pass. Product access and scenario tests were not run.
Failure status [unknown] Not established for failure status in the named first-party source pass. Product access and scenario tests were not run.
Approval status [supported] Tool-decision events identify approval or denial and whether configuration or user made the decision. Sources: codex:gap07-otel.
Reconnect consistency [unknown] Not established for reconnect consistency in the named first-party source pass. Product access and scenario tests were not run.
Performance method [unknown] Not established for performance method in the named first-party source pass. Product access and scenario tests were not run.
Responsiveness [unknown] Not established for responsiveness in the named first-party source pass. Product access and scenario tests were not run.
Per-task API and compute cost: turn.completed (TypeScript) and TurnResult (Python) report token usage. Not established: task cost, session cost, child cost, epic cost, harness aggregation, billed vs estimated, subscription allowance, provider reconciliation, compute cost, refresh delay, alerts, export, drilldown.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: task_cost, session_cost, child_cost, epic_cost, harness_aggregation, billed_vs_estimated, subscription_allowance, provider_reconciliation, compute_cost, refresh_delay, alerts, export, drilldown. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
14 atomic answers
Task cost [unknown] Not established for task cost in the named first-party source pass. Product access and scenario tests were not run.
Session cost [unknown] Not established for session cost in the named first-party source pass. Product access and scenario tests were not run.
Child cost [unknown] Not established for child cost in the named first-party source pass. Product access and scenario tests were not run.
Epic cost [unknown] Not established for epic cost in the named first-party source pass. Product access and scenario tests were not run.
Harness aggregation [unknown] Not established for harness aggregation in the named first-party source pass. Product access and scenario tests were not run.
Billed vs estimated [unknown] Not established for billed vs estimated in the named first-party source pass. Product access and scenario tests were not run.
Subscription allowance [unknown] Not established for subscription allowance in the named first-party source pass. Product access and scenario tests were not run.
Provider reconciliation [unknown] Not established for provider reconciliation in the named first-party source pass. Product access and scenario tests were not run.
Compute cost [unknown] Not established for compute cost in the named first-party source pass. Product access and scenario tests were not run.
Refresh delay [unknown] Not established for refresh delay in the named first-party source pass. Product access and scenario tests were not run.
Alerts [unknown] Not established for alerts in the named first-party source pass. Product access and scenario tests were not run.
Export [unknown] Not established for export in the named first-party source pass. Product access and scenario tests were not run.
Drilldown [unknown] Not established for drilldown in the named first-party source pass. Product access and scenario tests were not run.
Compute-aware scheduling and verification: agent_max_threads defaults to 6 and is configurable. Interrupt stops Code Mode cells and nested tools, and dropped remote MCP requests are cancelled. Not established: queues, cpu ram, backpressure, host awareness, check reuse, check deduplication, bounded tests, fairness, recovery, shared limits.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: queues, cpu_ram, backpressure, host_awareness, check_reuse, check_deduplication, bounded_tests, fairness, recovery, shared_limits. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
12 atomic answers
Concurrency [partial] agent_max_threads defaults to 6 and is configurable. Sources: codex:b10c-agent-config.
Queues [unknown] Not established for queues in the named first-party source pass. Product access and scenario tests were not run.
Cpu ram [unknown] Not established for cpu ram in the named first-party source pass. Product access and scenario tests were not run.
Backpressure [unknown] Not established for backpressure in the named first-party source pass. Product access and scenario tests were not run.
Host awareness [unknown] Not established for host awareness in the named first-party source pass. Product access and scenario tests were not run.
Check reuse [unknown] Not established for check reuse in the named first-party source pass. Product access and scenario tests were not run.
Check deduplication [unknown] Not established for check deduplication in the named first-party source pass. Product access and scenario tests were not run.
Bounded tests [unknown] Not established for bounded tests in the named first-party source pass. Product access and scenario tests were not run.
Cancellation [supported] Interrupt stops Code Mode cells and nested tools, and dropped remote MCP requests are cancelled. Sources: codex:b10c-rel-0163, codex:b10c-rel-0165.
Fairness [unknown] Not established for fairness in the named first-party source pass. Product access and scenario tests were not run.
Recovery [unknown] Not established for recovery in the named first-party source pass. Product access and scenario tests were not run.
Shared limits [unknown] Not established for shared limits in the named first-party source pass. Product access and scenario tests were not run.
Business includes SAML SSO; Enterprise adds SCIM and RBAC. Workspace switching checks membership before switching (experimental API).
Scope
Surfaces: Codex CLI, Codex cloud, Codex in ChatGPT desktop, Codex IDE extension · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: project_roles, task_roles, provisioning, unauthorized_access, revocation, plan_scope. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
9 atomic answers
Organization roles [supported] Workspaces have built-in Owner, Admin, Member and Analytics Viewer roles, plus custom roles that grant Codex local, Codex cloud and shared-environment permissions. Sources: codex:src2-codex-roles.
Project roles [unknown] Not established for project roles in the named first-party source pass. Product access and scenario tests were not run.
Task roles [unknown] Not established for task roles in the named first-party source pass. Product access and scenario tests were not run.
Sso [supported] Business includes SAML SSO; Enterprise adds SCIM and RBAC. Sources: codex:nov07-pricing.
Provisioning [unknown] Not established for provisioning in the named first-party source pass. Product access and scenario tests were not run.
Unauthorized access [unknown] Not established for unauthorized access in the named first-party source pass. Product access and scenario tests were not run.
Revocation [unknown] Not established for revocation in the named first-party source pass. Product access and scenario tests were not run.
Plan scope [unknown] Not established for plan scope in the named first-party source pass. Product access and scenario tests were not run.
Local approval policies control escalation beyond the sandbox; /permissions exposes read-only and other configurations. Network permissions and approval policy are separate controls. Remote projects use SSH; device control uses an authorized secure relay rather than exposing the host publicly. A read-only sandbox mode is documented for local Codex use. MCP tools have per-tool approval modes, and approval prompts include tool names and annotation hints. Managed checkpoints require a human decision that Guardian automatic review cannot satisfy; approver identity is not recorded in the sources read. Session and persistent approvals exist and cannot satisfy managed checkpoints. Exec-policy rules allow commands, and codex execpolicy check tests them. Granular policy auto-rejects categories, and approved commands keep explicit filesystem denials. The default approval policy is on-request. The OS sandbox and network proxy enforce boundaries locally; no bypass test was run.
Scope
Surfaces: Codex CLI, Codex IDE extension · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: secret, delivery, revoke. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
14 atomic answers
Command [supported] Local approval policies control escalation beyond the sandbox; /permissions exposes read-only and other configurations. Sources: codex:nov07-permissions.
Tool [supported] MCP tools have per-tool approval modes, and approval prompts include tool names and annotation hints. Sources: codex:b10c-mcp-tools, codex:b10c-rel-0167.
Network [supported] Network permissions and approval policy are separate controls. Sources: codex:nov07-permissions.
Secret [unknown] Not established for secret in the named first-party source pass. Product access and scenario tests were not run.
Delivery [unknown] Not established for delivery in the named first-party source pass. Product access and scenario tests were not run.
Approver identity [partial] Managed checkpoints require a human decision that Guardian automatic review cannot satisfy; approver identity is not recorded in the sources read. Sources: codex:b10c-as-threads.
Lifetime [partial] Session and persistent approvals exist and cannot satisfy managed checkpoints. Sources: codex:b10c-as-threads.
Revoke [unknown] Not established for revoke in the named first-party source pass. Product access and scenario tests were not run.
Tunnels [supported] Remote projects use SSH; device control uses an authorized secure relay rather than exposing the host publicly. Sources: codex:gap07-remote.
Defaults [supported] The default approval policy is on-request. Sources: codex:b10c-approval.
Safer modes [supported] A read-only sandbox mode is documented for local Codex use. Sources: codex:nov07-permissions.
Enforcement [partial] The OS sandbox and network proxy enforce boundaries locally; no bypass test was run. Sources: codex:b10c-sandbox.
Opt-in OTel tool-decision events retain approved or denied decisions and decision source. OTel exports selected events to user-controlled collectors. Saved local session rollouts record agent actions. Saved sessions persist across runs. Threads can be archived, deleted or reverted.
Scope
Surfaces: Codex CLI, Codex IDE extension, Codex in ChatGPT desktop, Codex cloud · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: First-party documentation; client, plan and execution restrictions are named per atomic answer. Unlisted parity is unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: edits, discussion, decisions, cross_user_persistence, retention, integrity. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
11 atomic answers
Actions [partial] Saved local session rollouts record agent actions. Sources: codex:b10c-rollout.
Approvals [supported] Opt-in OTel tool-decision events retain approved or denied decisions and decision source. Sources: codex:gap07-otel.
Edits [unknown] Not established for edits in the named first-party source pass. Product access and scenario tests were not run.
Discussion [unknown] Not established for discussion in the named first-party source pass. Product access and scenario tests were not run.
Decisions [unknown] Not established for decisions in the named first-party source pass. Product access and scenario tests were not run.
Cross user persistence [unknown] Not established for cross user persistence in the named first-party source pass. Product access and scenario tests were not run.
Review experience and guidance: Patch details expand in the transcript during approval. codex review reports on uncommitted changes, a base branch or a commit. Keyboard transcript navigation, a configurable leader key and keymap. Not established: pending reviews, evidence navigation, comments, request changes, effort, missed evidence, narrow screen, context switching, roles, retained context.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: pending_reviews, evidence_navigation, comments, request_changes, effort, missed_evidence, narrow_screen, context_switching, roles, retained_context. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
13 atomic answers
Pending reviews [unknown] Not established for pending reviews in the named first-party source pass. Product access and scenario tests were not run.
Change understanding [partial] Patch details expand in the transcript during approval. Sources: codex:b10c-rel-0167.
Evidence navigation [unknown] Not established for evidence navigation in the named first-party source pass. Product access and scenario tests were not run.
Comments [unknown] Not established for comments in the named first-party source pass. Product access and scenario tests were not run.
Request changes [unknown] Not established for request changes in the named first-party source pass. Product access and scenario tests were not run.
Shipping guidance [partial] codex review reports on uncommitted changes, a base branch or a commit. Sources: codex:b10c-exec-cli.
Effort [unknown] Not established for effort in the named first-party source pass. Product access and scenario tests were not run.
Missed evidence [unknown] Not established for missed evidence in the named first-party source pass. Product access and scenario tests were not run.
Keyboard [supported] Keyboard transcript navigation, a configurable leader key and keymap. Sources: codex:b10c-rel-0163.
Narrow screen [unknown] Not established for narrow screen in the named first-party source pass. Product access and scenario tests were not run.
Context switching [unknown] Not established for context switching in the named first-party source pass. Product access and scenario tests were not run.
Roles [unknown] Not established for roles in the named first-party source pass. Product access and scenario tests were not run.
Retained context [unknown] Not established for retained context in the named first-party source pass. Product access and scenario tests were not run.
Cloud review shows changed files and check results. Viewed and generated images preview in the transcript, and stored thread attachments are managed through app-server. Local rollout files keep thread items; cloud retention was not established. wait_agent can return a subagent's final message to its parent.
Scope
Surfaces: Codex cloud, Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: managed-cloud · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: documents, recordings, previews, test_outputs, reviewer_outputs, gallery, task_links, revision_links, prompt_leakage. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
Documents [unknown] Not established for documents in the named first-party source pass. Product access and scenario tests were not run.
Screenshots [partial] Viewed and generated images preview in the transcript, and stored thread attachments are managed through app-server. Sources: codex:b10c-rel-0164, codex:b10c-as-threads.
Recordings [unknown] Not established for recordings in the named first-party source pass. Product access and scenario tests were not run.
Previews [unknown] Not established for previews in the named first-party source pass. Product access and scenario tests were not run.
Test outputs [unknown] Not established for test outputs in the named first-party source pass. Product access and scenario tests were not run.
Reviewer outputs [unknown] Not established for reviewer outputs in the named first-party source pass. Product access and scenario tests were not run.
Gallery [unknown] Not established for gallery in the named first-party source pass. Product access and scenario tests were not run.
Task links [unknown] Not established for task links in the named first-party source pass. Product access and scenario tests were not run.
Revision links [unknown] Not established for revision links in the named first-party source pass. Product access and scenario tests were not run.
Persistence [partial] Local rollout files keep thread items; cloud retention was not established. Sources: codex:b10c-rollout.
Child findings [partial] wait_agent can return a subagent's final message to its parent. Sources: codex:b10c-multi-agents.
Prompt leakage [unknown] Not established for prompt leakage in the named first-party source pass. Product access and scenario tests were not run.
A dedicated local review can inspect uncommitted changes, a commit or a base branch and return prioritized findings. The stable, default-on shell tool runs test commands inside the sandbox; no dedicated test gate was found.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: local · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: lint, types, ci, spec_checks, required_checks, failure_blocking, requirement_conflicts, configured_policy, enforced_policy. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
11 atomic answers
Tests [partial] The stable, default-on shell tool runs test commands inside the sandbox; no dedicated test gate was found. Sources: codex:b10c-features.
Lint [unknown] Not established for lint in the named first-party source pass. Product access and scenario tests were not run.
Types [unknown] Not established for types in the named first-party source pass. Product access and scenario tests were not run.
Ci [unknown] Not established for ci in the named first-party source pass. Product access and scenario tests were not run.
Independent review [supported] A dedicated local review can inspect uncommitted changes, a commit or a base branch and return prioritized findings. Sources: codex:nov07-cli.
Spec checks [unknown] Not established for spec checks in the named first-party source pass. Product access and scenario tests were not run.
Required checks [unknown] Not established for required checks in the named first-party source pass. Product access and scenario tests were not run.
Failure blocking [unknown] Not established for failure blocking in the named first-party source pass. Product access and scenario tests were not run.
Requirement conflicts [unknown] Not established for requirement conflicts in the named first-party source pass. Product access and scenario tests were not run.
Configured policy [unknown] Not established for configured policy in the named first-party source pass. Product access and scenario tests were not run.
Enforced policy [unknown] Not established for enforced policy in the named first-party source pass. Product access and scenario tests were not run.
Cloud results can be committed or opened as pull requests after review. Cloud tasks take a branch, and local reviews compare against a base branch. Cloud users can request follow-up changes in the same chat; pull-request review comments were not established.
Scope
Surfaces: Codex cloud, Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: managed-cloud · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: commits, ci_response, rebase, conflicts, product_state. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
Worktrees [supported] The CLI's --worktree option runs a session in a new managed Git worktree (released in rust-v0.167.0). Sources: codex:src2-codex-worktree-flag.
Commits [unknown] Not established for commits in the named first-party source pass. Product access and scenario tests were not run.
Pull requests [supported] Cloud results can be committed or opened as pull requests after review. Sources: codex:nov07-cloud.
Comments [partial] Cloud users can request follow-up changes in the same chat; pull-request review comments were not established. Sources: codex:b10c-cloud-doc.
Ci response [unknown] Not established for ci response in the named first-party source pass. Product access and scenario tests were not run.
Rebase [unknown] Not established for rebase in the named first-party source pass. Product access and scenario tests were not run.
Conflicts [unknown] Not established for conflicts in the named first-party source pass. Product access and scenario tests were not run.
Product state [unknown] Not established for product state in the named first-party source pass. Product access and scenario tests were not run.
Completion evidence and rollback: thread/revert rewinds a thread in the app-server; code rollback was not established. Not established: commit links, pr links, build links, deployment links, delivery status, retained evidence, rollback, stale approval invalidation.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: commit_links, pr_links, build_links, deployment_links, delivery_status, retained_evidence, rollback, stale_approval_invalidation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
9 atomic answers
Commit links [unknown] Not established for commit links in the named first-party source pass. Product access and scenario tests were not run.
Pr links [unknown] Not established for pr links in the named first-party source pass. Product access and scenario tests were not run.
Build links [unknown] Not established for build links in the named first-party source pass. Product access and scenario tests were not run.
Deployment links [unknown] Not established for deployment links in the named first-party source pass. Product access and scenario tests were not run.
Delivery status [unknown] Not established for delivery status in the named first-party source pass. Product access and scenario tests were not run.
Retained evidence [unknown] Not established for retained evidence in the named first-party source pass. Product access and scenario tests were not run.
Rollback [unknown] Not established for rollback in the named first-party source pass. Product access and scenario tests were not run.
Revert [partial] thread/revert rewinds a thread in the app-server; code rollback was not established. Sources: codex:b10c-as-threads.
Stale approval invalidation [unknown] Not established for stale approval invalidation in the named first-party source pass. Product access and scenario tests were not run.
The repository LICENSE at fc01c073b9ea09c06347a234a1ca231a9ef65bf5 is Apache-2.0; hosted components and separate clients require their own licensing checks. The IDE extension and Codex Cloud are explicitly not open source. CLI, SDK and App Server sources are public; IDE extension and cloud service are not open source. Public CLI/App Server source does not establish hosted Cloud or IDE implementation parity. Latest stable release rust-v0.171.0 (2026-11-20) maps to commit ccf9be8.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: server_license. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
6 atomic answers
Client license [supported] The repository LICENSE at fc01c073b9ea09c06347a234a1ca231a9ef65bf5 is Apache-2.0; hosted components and separate clients require their own licensing checks. Sources: codex:nov07-license.
Server license [unknown] Not established for server license in the named first-party source pass. Product access and scenario tests were not run.
Hosted extensions [supported] The IDE extension and Codex Cloud are explicitly not open source. Sources: codex:gap07-open-source.
Source visibility [supported] CLI, SDK and App Server sources are public; IDE extension and cloud service are not open source. Sources: codex:gap07-open-source.
The Codex CLI and App Server core can run on user-controlled infrastructure under the public source distribution. This hosts the coding loop and local session/control process, while inference uses its configured provider. Codex Cloud and the closed IDE/desktop components are not claimed self-hostable. Sessions are local JSONL rollout files under CODEX_HOME; a hosted export was not found. A TUI flow migrates configuration from other coding agents; importing their memories is behind an off-by-default flag under development. JSONL session rollouts and JSONL exec events. The Ollama and LM Studio providers need no OpenAI sign-in; this was not tested.
Scope
Surfaces: Codex CLI · Version: Earlier findings: Codex rust-v0.167.0 source pinned at 086ccc833bc9cf01024ad2dd6af7dbd421b6bf9a and current first-party docs checked 2026-11-13. No installed release tested.; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Not specified · OS: Not specified · Execution modes: user-controlled CLI and App Server · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-13 (9 days old)
Agent cross-check
2026-11-22
Only the named documented or inspected parts are established; other atomic questions retain their recorded gaps.
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: cancellation, history_portability. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
8 atomic answers
Self hosting [supported] The native CLI runs the core coding agent and can start/manage its local App Server daemon on user-controlled machines. Sources: codex:gap-self-hosted-core.
Vendor dependencies [partial] The local core still requires configured model inference/authentication; this does not self-host Codex Cloud. Sources: codex:gap-self-hosted-core.
Export [partial] Sessions are local JSONL rollout files under CODEX_HOME; a hosted export was not found. Sources: codex:b10c-rollout.
Import [partial] A TUI flow migrates configuration from other coding agents; importing their memories is behind an off-by-default flag under development. Sources: codex:b10c-ext-migration, codex:b10c-features.
Cancellation [unknown] Not established for cancellation in the named first-party source pass. Product access and scenario tests were not run.
Account independence [partial] The Ollama and LM Studio providers need no OpenAI sign-in; this was not tested. Sources: codex:b10c-providers.
History portability [unknown] Not established for history portability in the named first-party source pass. Product access and scenario tests were not run.
The Business plan states no training on business data by default; consumer routes are not covered by this statement. Built-in Statsig analytics and independently configured OTel exporters have distinct controls. The default Statsig metrics exporter follows analytics.enabled; the reference says unset analytics uses the client default. OTel log export is separately opt-in. Set analytics.enabled=false for built-in Statsig metrics, and set OTel log, metrics and trace exporters to none to disable explicitly configured telemetry exports. OTel prompt and response logging default to off and need opt-in. Credentials can be stored in a file, the OS keyring or keyring with fallback, and are masked in supported remote shells. ChatGPT workspace routing reports us, us_cr or NO_CONSTRAINT.
Scope
Surfaces: Codex CLI, Codex cloud, Codex in ChatGPT desktop, Codex IDE extension · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: code_data, logs, artifacts, retention, deletion, subprocessors, consent_prompt, provider_separation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
15 atomic answers
Code data [unknown] Not established for code data in the named first-party source pass. Product access and scenario tests were not run.
Prompts [partial] OTel prompt and response logging default to off and need opt-in. Sources: codex:b10c-cred-store.
Credentials [partial] Credentials can be stored in a file, the OS keyring or keyring with fallback, and are masked in supported remote shells. Sources: codex:b10c-cred-store, codex:b10c-rel-0164.
Logs [unknown] Not established for logs in the named first-party source pass. Product access and scenario tests were not run.
Artifacts [unknown] Not established for artifacts in the named first-party source pass. Product access and scenario tests were not run.
Retention [unknown] Not established for retention in the named first-party source pass. Product access and scenario tests were not run.
Deletion [unknown] Not established for deletion in the named first-party source pass. Product access and scenario tests were not run.
Training [supported] The Business plan states no training on business data by default; consumer routes are not covered by this statement. Sources: codex:nov07-pricing.
Subprocessors [unknown] Not established for subprocessors in the named first-party source pass. Product access and scenario tests were not run.
Telemetry disclosure [supported] Built-in Statsig analytics and independently configured OTel exporters have distinct controls. Sources: codex:gap07-config.
Telemetry defaults [partial] The default Statsig metrics exporter follows analytics.enabled; the reference says unset analytics uses the client default. OTel log export is separately opt-in. Sources: codex:gap07-config.
Telemetry controls [supported] Set analytics.enabled=false for built-in Statsig metrics, and set OTel log, metrics and trace exporters to none to disable explicitly configured telemetry exports. Sources: codex:gap07-config.
Consent prompt [unknown] Not established for consent prompt in the named first-party source pass. Product access and scenario tests were not run.
Provider separation [unknown] Not established for provider separation in the named first-party source pass. Product access and scenario tests were not run.
Codex offers Free and recurring paid ChatGPT plans, with separate API-key billing. Plus/Pro users can purchase additional credits after included limits, and eligible workspace plans can buy shared credits. Existing USD plan prices retain their recorded scope.
Scope
Surfaces: Codex CLI, Codex cloud, Codex in ChatGPT desktop, Codex IDE extension · Version: Codex rust-v0.167.0 source pinned at 086ccc833bc9cf01024ad2dd6af7dbd421b6bf9a and current first-party docs checked 2026-11-13. No installed release tested. · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-10 (12 days old)
Agent cross-check
Not separately cross-checked
Only the named documented or inspected parts are established; other atomic questions retain their recorded gaps.
12 atomic answers
Currency [supported] Published prices are USD. Sources: codex:nov10-pricing.
Interval [supported] Go costs $8/month and Plus $20/month; Pro starts at $100/month. Business costs $20/user/month billed annually or $25/user/month billed monthly. Sources: codex:nov10-pricing.
Seats [supported] Business is priced per user; a minimum is stated only for the annual rate. Sources: codex:nov10-pricing.
Seat minimum [supported] Business annual pricing requires at least two users. Sources: codex:nov10-pricing.
Workspace fees [unknown] Not established for workspace fees in the named first-party source pass. Product access and scenario tests were not run.
Included usage [partial] Free and Go include desktop-app usage subject to rollout; Plus includes web, CLI, IDE and iOS, with model-dependent shared usage. Sources: codex:nov10-pricing.
Credits [supported] Users can purchase additional personal or workspace credits under the documented eligible plans. Sources: codex:gap-credit-billing.
Trial [unknown] Not established for trial in the named first-party source pass. Product access and scenario tests were not run.
Overages [supported] Additional credits or separately billed API-key usage extend use beyond included plan limits. Sources: codex:gap-credit-billing.
Taxes [unknown] Not established for taxes in the named first-party source pass. Product access and scenario tests were not run.
Model cost [supported] API-key usage follows separate API pricing; API rates do not determine included subscription task counts. Sources: codex:nov10-pricing.
Compute cost [unknown] Not established for compute cost in the named first-party source pass. Product access and scenario tests were not run.
The Linux desktop client is in preview, with named x64/ARM64 distribution packages. Linux preview supports Ubuntu 24.04/26.04, Debian 13, Fedora 43/44 and current Arch. Public GitHub issues with per-surface templates, and a Bugcrowd security channel; no response commitment. Stable releases from 0.157.0 (2026-09-25) to 0.171.0 (2026-11-20), with 0.173.0 alphas through 2026-11-21. Release code rejects the removed ollama-chat provider and names ollama as the replacement. The docs label Codex Cloud (Legacy) and point to a newer guide, without an end date.
Scope
Surfaces: Codex CLI, Codex IDE extension, Codex in ChatGPT desktop, Codex cloud · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: First-party documentation; client, plan and execution restrictions are named per atomic answer. Unlisted parity is unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
unknown
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: response_commitments, status_history, export_options. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
8 atomic answers
Release stage [supported] The Linux desktop client is in preview, with named x64/ARM64 distribution packages. Sources: codex:gap07-linux.
Access restrictions [supported] Linux preview supports Ubuntu 24.04/26.04, Debian 13, Fedora 43/44 and current Arch. Sources: codex:gap07-linux.
Support [partial] Public GitHub issues with per-surface templates, and a Bugcrowd security channel; no response commitment. Sources: codex:b10c-security.
Response commitments [unknown] Not established for response commitments in the named first-party source pass. Product access and scenario tests were not run.
Release activity [supported] Stable releases from 0.157.0 (2026-09-25) to 0.171.0 (2026-11-20), with 0.173.0 alphas through 2026-11-21. Sources: codex:b10c-releases.
Status history [unknown] Not established for status history in the named first-party source pass. Product access and scenario tests were not run.
Deprecation [partial] Release code rejects the removed ollama-chat provider and names ollama as the replacement. The docs label Codex Cloud (Legacy) and point to a newer guide, without an end date. Sources: codex:b10c-providers, codex:b10c-cloud-env.
Export options [unknown] Not established for export options in the named first-party source pass. Product access and scenario tests were not run.
A $0 monthly plan includes limited desktop Codex usage, subject to rollout. Free access is limited to listed desktop model/speed eligibility; detailed task allowances depend on the plan and model. The Apache-2.0 CLI has no license fee; with Ollama or LM Studio inference runs locally. Not tested.
Scope
Surfaces: Codex CLI, Codex cloud, Codex in ChatGPT desktop, Codex IDE extension · Version: Earlier findings: Public sources checked 2026-11-07; no installed release tested; 2026-11-22 inherited-lead additions: openai/codex rust-v0.171.0 source at ccf9be8 (2026-11-19; release 2026-11-20), release notes rust-v0.157.0 to rust-v0.171.0 (2026-09-25 to 2026-11-20) and learn.chatgpt.com cloud, cloud-environment and pricing pages read 2026-11-22 · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-22
Unknown after the 2026-11-22 inherited-lead recheck of first-party docs, releases and public code: trial, credits_trial, paid_only, early_access, free_scenarios, card_required, paid_evaluation. No configured product fixture was run; documentation does not establish successful hands-on behavior.
Documentation, release and public-code recheck only (2026-11-22); no configured product fixture was run.
10 atomic answers
Ongoing free [supported] A $0 monthly plan includes limited desktop Codex usage, subject to rollout. Sources: codex:gap07-pricing.
Trial [unknown] Not established for trial in the named first-party source pass. Product access and scenario tests were not run.
Credits trial [unknown] Not established for credits trial in the named first-party source pass. Product access and scenario tests were not run.
Self host cost [partial] The Apache-2.0 CLI has no license fee; with Ollama or LM Studio inference runs locally. Not tested. Sources: codex:b10c-license, codex:b10c-providers.
Paid only [unknown] Not established for paid only in the named first-party source pass. Product access and scenario tests were not run.
Early access [unknown] Not established for early access in the named first-party source pass. Product access and scenario tests were not run.
Free scenarios [unknown] Not established for free scenarios in the named first-party source pass. Product access and scenario tests were not run.
Card required [unknown] Not established for card required in the named first-party source pass. Product access and scenario tests were not run.
Usage caps [partial] Free access is limited to listed desktop model/speed eligibility; detailed task allowances depend on the plan and model. Sources: codex:gap07-pricing.
Paid evaluation [unknown] Not established for paid evaluation in the named first-party source pass. Product access and scenario tests were not run.
Prices distinguish individual subscriptions, Business seats and separate API usage. Enterprise and Edu require contacting sales.
Scope
Surfaces: Codex CLI, Codex cloud, Codex in ChatGPT desktop, Codex IDE extension · Version: Public sources checked 2026-11-07; no installed release tested · Plan: Plan restrictions and client/mode boundaries are stated per atomic answer; unlisted parity remains unknown. · OS: Not specified · Execution modes: Not specified · Region: Not specified · Harness: Not specified
Availability
unknown
Delivery
native
Checked
2026-11-07 (15 days old)
Agent cross-check
2026-11-07
Pricing transparency: first-party sources in this pass did not establish included_limits, overages, compute_markup, model_markup, minimum_commitment, cancellation, transparency_classification. No full license/hosted parity audit, assurance-report access or fixed one/two/five-person workload costing was performed. Public pricing is dated; missing rates remain unknown.
9 atomic answers
Units [supported] Prices distinguish individual subscriptions, Business seats and separate API usage. Sources: codex:nov07-pricing.
Included limits [unknown] Not established for included limits in the named first-party source pass. Product access and scenario tests were not run.
Overages [unknown] Not established for overages in the named first-party source pass. Product access and scenario tests were not run.
Compute markup [unknown] Not established for compute markup in the named first-party source pass. Product access and scenario tests were not run.
Model markup [unknown] Not established for model markup in the named first-party source pass. Product access and scenario tests were not run.
Minimum commitment [unknown] Not established for minimum commitment in the named first-party source pass. Product access and scenario tests were not run.
Cancellation [unknown] Not established for cancellation in the named first-party source pass. Product access and scenario tests were not run.
Sales only unknowns [supported] Enterprise and Edu require contacting sales. Sources: codex:nov07-pricing.
Transparency classification [unknown] Not established for transparency classification in the named first-party source pass. Product access and scenario tests were not run.
The desktop application offers Codex chats and downloads for macOS, Windows and Linux. This evidence concerns the Codex mode, not unrelated ChatGPT Work features.
Codex in the ChatGPT iOS app can start, steer, approve and review cloud or connected-computer tasks. Local execution needs its connected computer awake. Rollout and workspace policy affect access.
Codex supports iOS and Android control subject to rollout. SSH project chats execute on the remote filesystem after Codex is installed/authenticated there. Handoff moves a chat and Git state between hosts.
Kind
documented
Locator
Before mobile setup; Connect to an SSH host; Handoff
The current app is named ChatGPT desktop app and offers ChatGPT or Codex in its product selector. This establishes a shared desktop surface while retaining Codex as a choice; it does not announce a product-wide Codex rename or merger date.
Kind
documented
Locator
Get started with the desktop app > Send your first message; exact sentence: "Choose ChatGPT or Codex."
The comparison distinguishes ChatGPT Work and Codex in the same desktop app and directs users to select Codex in the product selector. Codex retains developer project/diff/review surfaces.
Kind
documented
Locator
Compare ChatGPT Work and Codex on desktop > Detailed comparison > Where to start
Codex CLI is the product-owned interactive coding agent that runs on local repositories, supports resumable chats and can submit cloud work. It is distinct from a model-provider adapter.
The CLI executes the Codex core and creates native Codex agent threads through AgentControl. Configuration selects model providers and roles, not an alternate external coding-harness executable. Bounded inspection of CLI/core/config and multi-agent handlers found no native Claude Code, Cursor, Gemini CLI, OpenCode, Pi, Grok, Aider or Antigravity adapter.
Free is $0 monthly for limited desktop usage subject to rollout. Plus $20 includes CLI, IDE, web and iOS. API-key use excludes cloud integrations. Business is $25/user monthly with at least two users; annual rate differs.
Built-in providers are OpenAI, Amazon Bedrock, Google Vertex (GCP), Ollama and LM Studio; the local OSS providers are created without requires_openai_auth. Other compatible providers can be added in config.toml. The old ollama-chat provider ID now fails with a "no longer supported" error telling users to switch to ollama.
Kind
source-inspected
Locator
built_in_model_providers; create_oss_provider; requires_openai_auth; LEGACY_OLLAMA_CHAT_PROVIDER_ID; tag rust-v0.171.0 (commit ccf9be8)
App-server clients can list a managed ChatGPT login's workspaces and switch with accountSession/switch; Core checks membership, stops loaded threads and validates routing (experimental API). account/read.workspaceRouting returns the selected workspace routing value us, us_cr or NO_CONSTRAINT. thread/list can be scoped to the current ChatGPT login.
Kind
documented
Locator
Workspace switching (experimental); account/read.workspaceRouting; Local history by account; tag rust-v0.171.0 (commit ccf9be8)
codex cloud exec submits a task to a named environment and an optional Git branch (default the current branch); status, list (filter by environment), diff and apply inspect and apply results. Signed-out users are told to run codex login with ChatGPT before codex cloud.
Kind
source-inspected
Locator
enum Command (Exec, Status, List, Apply, Diff); ExecCommand --env and --branch; lib.rs "Not signed in" message; tag rust-v0.171.0 (commit ccf9be8)
Codex worktree chats isolate repository working files and can be handed between Local and Worktree. Execution remains on the connected project computer or its remote development environment.
Cloud tasks can use repositories from a connected GitHub or GitLab account; users review changed files and checks, request follow-up changes in the same chat, and commit or open a pull request.
Credentials can be stored in a file under CODEX_HOME, the OS keyring, or keyring with file fallback. OTel prompt and agent-response logging default to false. TUI desktop notifications are configurable and default to on.
Kind
source-inspected
Locator
AuthCredentialsStoreMode (File, Keyring, Auto); OtelConfig log_user_prompt and log_agent_responses defaults; tag rust-v0.171.0 (commit ccf9be8)
@openai/codex-sdk (Node.js 18+) spawns the codex CLI and exchanges JSONL events; threads are started, run, continued and streamed, and turn.completed events report usage.
Kind
documented
Locator
Installation; Quickstart; streaming events; tag rust-v0.171.0 (commit ccf9be8)
The openai-codex Python package starts threads, runs turns and streams progress; thread.run returns a TurnResult with the final response, items and token usage.
Kind
documented
Locator
Install; Quickstart; tag rust-v0.171.0 (commit ccf9be8)
The experimental app-server daemon backs the app-server lifecycle commands used by remote clients such as the desktop and mobile apps, on Linux, macOS and Windows.
Kind
documented
Locator
intro; Platform support; tag rust-v0.171.0 (commit ccf9be8)
Skills are SKILL.md folders loaded from skills directories, including project .agents/skills locations, plus bundled system skills. Lifecycle hooks cover session start/end, user prompt submit, pre/post tool use, permission requests, model requests, compaction, interrupt and stop.
Kind
source-inspected
Locator
skills/src/lib.rs SKILLS_DIR_NAME, SYSTEM_SKILLS_DIR_NAME; ext/skills/src/host_roots.rs AGENTS_DIR_NAME ".agents" and repo_agents_skill_roots (repository .agents/skills roots); hooks/src/events; tag rust-v0.171.0 (commit ccf9be8)
Project instructions are AGENTS.md files collected from the project root down to the working directory, with AGENTS.override.md as a preferred local override.
Kind
source-inspected
Locator
module docs; DEFAULT AGENTS.md filename; LOCAL_AGENTS_MD_FILENAME; tag rust-v0.171.0 (commit ccf9be8)
shell_tool, guardian_approval and goals are Stable and on by default. memories is Stable but off by default. external_agent_memory_import is Under Development and off. The external_migration flag is Removed (kept only for backward compatibility).
The TUI migration flow imports configuration from other coding agents (Claude- and Cursor-style sources), including hooks, MCP servers, plugins, subagents and, where the source supports it, memory.
Enterprise MCP can authenticate with Codex account grants. Credentials are masked in supported remote Bash and Zsh sessions. Viewed and generated images preview in the transcript. Failed compaction leaves no partial summaries.
The command center and session picker mark unread results. Patch details expand during approval. Changelog: MCP approval metadata includes tool names and annotation hints (#58161); transient environment startup failures retry on the next turn (#57866); subagent relationships persist (#58129).
The public repository contains only a launcher that opens or installs the Desktop app on macOS and Windows; the README points to separately distributed IDE extensions and the desktop app. The desktop and IDE client code is not in openai/codex; the CLI/TUI is Rust.
Kind
source-inspected
Locator
mod.rs run_app_open_or_install (mac.rs, windows.rs); README IDE and desktop links; tag rust-v0.171.0 (commit ccf9be8)
codex exec takes a prompt argument or stdin (piped stdin is appended), optional images, JSONL event output and an output schema. codex review checks uncommitted changes, a base branch or a commit.
Kind
source-inspected
Locator
prompt/stdin handling; --image; --json; --output-schema; ReviewArgs; tag rust-v0.171.0 (commit ccf9be8)
A thread goal has an objective, a status (active, paused, blocked, usage-limited, budget-limited, complete), an optional token budget and tokens/time used; goals persist in the local state database.
Kind
source-inspected
Locator
protocol.rs ThreadGoal, ThreadGoalStatus; state/goals_migrations/0001_thread_goals.sql (the earlier state/migrations/0029_thread_goals.sql was dropped by 0034_drop_thread_goals.sql); tag rust-v0.171.0 (commit ccf9be8)
The model gets tools to spawn agents with an optional role, send messages or follow-up tasks to existing agents, resume closed agents, wait for final status (which may include the final message), list live agents in the thread tree, and close or interrupt agents.
The session picker gets a repository filter and the command center directory grouping across linked worktrees. Reasoning effort can change mid-conversation by default. Memory v2 is the default for fresh profiles. Dropped remote MCP tool requests are cancelled.
The memory pipeline version setting says fresh profiles default to v2 when memories are enabled (unless dual-write is explicitly disabled), and existing v1 profiles default to v1.
Kind
source-inspected
Locator
struct MemoriesToml, field version doc comment; tag rust-v0.171.0 (commit ccf9be8)
update_plan supplies a per-turn checklist of step/status values and emits it into one thread event stream. The app-server sends turn/plan/updated notifications. Bounded source inspection found no separate user-managed cross-session issue store; hosted Cloud/desktop task management is not public source.
Plus and Business show estimated local messages per five-hour period, Pro has no five-hour limit, cloud tasks may use more allowance, and weekly limits may apply; these are estimates, not fixed limits.
The five-hour estimates cover Plus and Standard Business; Pro plans currently have no five-hour limit; weekly limits may also apply; Enterprise/Edu users with flexible pricing have no fixed rate limits and usage scales with credits, while Enterprise and Edu plans without flexible pricing have the same per-seat limits as Plus for most features.
Kind
documented
Locator
usage-limit estimates introduction and footnotes below the table
Desktop Codex tasks run on a schedule locally or in worktrees; the app and computer must remain running. Scheduled shows findings and unread attention. CLI and IDE have no Scheduled management UI.
Managed requirements.toml prompt-user rules need a fresh human decision that automatic review and session or persistent approvals cannot satisfy; app-server support is experimental. Stored threads carry durable attachments managed through thread/attachment/* methods. thread/rollback was removed in favour of thread/revert, which rewinds a thread. Threads can be archived and deleted.
Kind
documented
Locator
Managed human command checkpoints; Stored thread attachments; Thread rollback (thread/revert); thread/archive and thread/delete; tag rust-v0.171.0 (commit ccf9be8)
An optional notify command is spawned for end-user notifications, and the TUI sends desktop notifications through terminal escape sequences, on by default.
Kind
source-inspected
Locator
notify; TuiNotificationSettings (types.rs); tui/src/notifications/osc.rs; tag rust-v0.171.0 (commit ccf9be8)
Unsent queued messages resume after reconnection once uncertain submissions are resolved, avoiding duplicate sends. Subagents receive configuration or preparation failures. The command center can browse older tasks.
/config searches preferences. Follow-ups steer a running turn; interrupting stops Code Mode cells and nested tool calls. /daemon restarts the background server, warns about interruptible work and reconnects. /keymap configures a leader key.
Approval requests (sandbox escapes, blocked network access, MCP approval prompts, ARC escalations) are routed to the user by default; automatic review by a prompted subagent applies only when approvals_reviewer is set to auto_review.
Kind
source-inspected
Locator
enum ApprovalsReviewer (#[default] User; AutoReview serialized auto_review, alias guardian_subagent) and its doc comment; tag rust-v0.171.0 (commit ccf9be8)
Opt-in OTel logs include API, stream, approval and tool events, with prompt text redacted unless enabled. Codex documents a devcontainer security-boundary option with explicit credential-exfiltration limits.
The workspace-write sandbox policy limits writes to the working directory plus configured writable roots, and outbound network access is off unless network_access is set to true.
The current first-party guide explicitly says each cloud task runs in a VM, with plan-dependent vCPU, RAM and disk resources and a recoverable saved-state window. Enterprise can request larger VM specifications.
On Linux the sandbox uses bubblewrap with PID (and optionally network) namespaces plus seccomp; macOS uses Seatbelt; a local network proxy enforces allow/deny and limited modes.
Kind
source-inspected
Locator
linux-sandbox/src/bwrap.rs (--unshare-pid, --unshare-net), seccomp.rs; sandboxing/src/seatbelt.rs; network-proxy/README.md; tag rust-v0.171.0 (commit ccf9be8)
The page is titled Codex Cloud (Legacy), which still serves Code Review and the Linear and GitHub integrations, and points to a newer Cloud environments guide; no end date is given. Secrets are available only to setup scripts and are removed before the agent phase.
Plus is $20/month; Business is $20/user/month annually for two or more users or $25 monthly; API keys have separate billing and exclude cloud features.
ChatGPT workspaces have built-in Owner, Admin, Member and Analytics Viewer roles with described rights, plus custom roles that grant Codex permissions such as local Codex use, Codex cloud use and managing shared workspace environments.
Kind
documented
Locator
Built-in roles and custom roles sections, fetched 2026-11-10
Stable releases rust-v0.157.0 (2026-09-25) through rust-v0.171.0 (2026-11-20, marked Latest; tag commit ccf9be8), with 0.173.0 alphas through 2026-11-21.
The public Apache-2.0 CLI runs the core coding agent locally and can start/manage its own app-server daemon on a user machine. The local core owns sessions, filesystem tools and process execution; it is not only a worker for the hosted Cloud control plane. Inference still uses the configured model provider.
Kind
source-inspected
Locator
Local Codex CLI and app-server daemon commands; repository README and LICENSE
Free is $0/month and Go is $8/month, both with GPT-6 Luna in the desktop app subject to rollout. Plus is $20/month with Codex on the web, CLI, IDE extension and iOS. Pro starts at $100/month. Business is $20/user/month billed annually for two or more users, or $25/user/month billed monthly. API-key use is billed at API pricing and has no cloud-based features. Prices are in USD.
Kind
documented
Locator
Pricing options, Individual and Business / Enterprise cards
Plus/Pro users can purchase additional credits; eligible Business/Edu/Enterprise workspaces can buy workspace credits. API-key use is separately usage-billed. Published plans are recurring subscriptions rather than one-time application licenses.
Kind
documented
Locator
What happens when you hit usage limits; credit-based usage rates; API Key plan