OpenAI Codex release notes: CLI 0.142.0 makes agent work easier to control

OpenAI Codex release notes now have a stable CLI signal worth paying attention to: @openai/codex is at 0.142.0 as npm latest, and the GitHub release bundles several control features that make Codex easier to run in real workflows. Not because every switch should be turned on today. More because teams that let Codex work with plugins, web search, multiple agent threads, and remote environments need better ways to see, limit, and audit the work.

Codex CLI is OpenAI Codex's coding agent in the terminal: it can read, change, and run code in the workspace you give it. A coding agent is an AI assistant that works with files, commands, and tools instead of only answering in chat. An approval gate is the checkpoint where a human approves the next action before the agent continues.

Source: OpenAI Codex GitHub release rust-v0.142.0, @openai/codex npm registry

OpenAI Codex release notes: what CLI 0.142.0 changes

The useful change in 0.142.0 is not one big button. It is a bundle of operating controls:

• /usage can show and redeem earned usage-limit reset credits, with confirmation, retry, and refreshed availability states.
• /plugins organizes remote plugins into OpenAI Curated, Workspace, and Shared with me, while eligible turns can recommend relevant plugins.
• Rollout budgets can track token usage across agent threads, remind users about remaining budget, and abort when the budget is exhausted.
• App-server clients can configure multi-agent delegation as disabled, explicit-request-only, or proactive.
• An indexed web-search mode can allow live searches while restricting direct page access to server-approved URLs.
• Codex can receive UTC time reminders and query the current time when time support is enabled.

That has release-note value for Nordic automation teams because it moves Codex from "the agent ran something" toward "we can see which limits, plugins, budgets, and information sources applied".

Source: OpenAI Codex release notes for rust-v0.142.0

Two commands are most visible to regular Codex users

The two clearest TUI signals are /usage and /plugins. TUI means text user interface: the terminal view where the human types slash commands, chooses options, and reviews the agent's work.

/usage is now the entry point for token activity and earned usage-limit resets. The PR text also says existing views such as /usage daily, /usage weekly, and /usage cumulative remain. If a personal account has reset credits available, the flow should show loading, empty, confirmation, success, retry, and error states.

/plugins gets more structure: the remote catalog appears as OpenAI Curated, Workspace, and Shared with me. For teams building internal Codex workflows, that is more than a UI tidy-up. A plugin is a package that can give Codex instructions, tools, or integrations. When plugins are grouped by source, it is easier to see what is local, workspace-managed, or shared by someone else.

Source: /usage reset flow PR #28154, /usage copy/state PR #28793, /plugins catalog PR #26703, plugin recommendations PR #27704

Why this matters before Codex gets more responsibility

Most of 0.142.0 points in the same direction: Codex should be able to work longer, with more tools, without the team losing control. A good prompt is not enough. You need a small run map.

For a Hammer-style team, that map can be short:

• which plugins Codex may use in this workspace
• which token budget applies to agent threads and subagents
• whether multi-agent work should require explicit request or may start proactively
• whether web search should be cached, live, indexed, or disabled
• where approvals, logs, environment variables, secret managers, scoped API keys, and redaction belong

This is a Tool Forge question more than a prompt question. When the agent gets real tools, permissions, sources, budget, and logging should be as clear as the task itself.

Source: multi-agent mode PR #29324, indexed web search PR #28489, rollout budget PR #28746, rollout budget abort PR #28707

Short example: use the new Codex feature

Human step: update or open a Codex CLI environment where you are actually running 0.142.0. Use /usage to check usage state and /plugins to see which plugins are available before asking the agent to do real work. Slash commands are human TUI steps, not text to paste into the agent prompt.

Then paste this prompt into Codex:

Review this Codex workspace as a 0.142.0 operations check. Return a short run-control card: current plugin needs, proposed multi-agent delegation mode, token-budget reminder points, web-search mode, and the approval gates a human should review before the next run.

Good output should include:

• a short list of plugins that are needed, and which ones are not
• a clear choice between no, explicit, or proactive multi-agent delegation
• concrete budget and web-search settings to review, not just "be careful"
• human approval gates for risky commands, network access, secrets, and production changes

What to watch next

The same day also has 0.143.0-alpha prereleases in GitHub and npm, but their release bodies are very thin at the time of writing. That makes the stable 0.142.0 release the better signal for practical Codex release notes today. The next thing to watch is whether these controls show up as more fully documented workflows in OpenAI Codex docs, especially around team configuration, plugin governance, and web search.