LIVE
Google Introduces "Gemini 4 Carbon" as Next Step in Its Model Lineup28/10/26 · Google|DeepMind unveils Genie 4, a new generation of world model27/10/26 · Google DeepMind|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|Google's AI gives evasive answers about who won the 2020 US election28/10/26 · Google|NVIDIA introduces Nemotron 3.5 Transcribe, a promptable multilingual transcription model28/10/26 · NVIDIA|OpenAI and ServiceNow deepen partnership on enterprise workflows28/10/26 · OpenAI|Failed startups' data finds a second life training AI models28/10/26|Anthropic expands AI for Science program to 20 foundations and research institutions28/10/26 · Anthropic|Anthropic shares early results on using Claude to accelerate drug development27/10/26 · Anthropic|GEPA-Curator: Filtering Failures to Better Train LLM Agents27/10/26|Google Introduces "Gemini 4 Carbon" as Next Step in Its Model Lineup28/10/26 · Google|DeepMind unveils Genie 4, a new generation of world model27/10/26 · Google DeepMind|Anthropic files IPO prospectus, with risk disclosures taking center stage26/10/26 · Anthropic|Robot policy trained only in simulation matches top real-world systems26/10/26|OpenAI unveils GPT-6.1 Terra, a model built for computer use26/10/26 · OpenAI|Google's AI gives evasive answers about who won the 2020 US election28/10/26 · Google|NVIDIA introduces Nemotron 3.5 Transcribe, a promptable multilingual transcription model28/10/26 · NVIDIA|OpenAI and ServiceNow deepen partnership on enterprise workflows28/10/26 · OpenAI|Failed startups' data finds a second life training AI models28/10/26|Anthropic expands AI for Science program to 20 foundations and research institutions28/10/26 · Anthropic|Anthropic shares early results on using Claude to accelerate drug development27/10/26 · Anthropic|GEPA-Curator: Filtering Failures to Better Train LLM Agents27/10/26|
ToolRisingofficial publisher

shell, apply_patch (Codex)

Tool : A precise action the AI can call · requires codex

Atomic commands and patches, under dual sandbox and approval control.

The cleanest security model among code agents: two orthogonal settings.

Overall score
7.9 /10
Security: serious reservations
Utility
9.0
Reliability
10.0
Security
Serious reservations
Ease
5 min · Comfortable

Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.

What it does, concretely

Working with code from a terminal usually means switching between reading files, editing them manually, and running commands. Codex CLI brings these three actions into a single loop driven by a language model, without leaving the terminal.

Concretely, the tool exposes two primitives: the shell command, which executes system instructions on the local machine, and apply_patch, which applies file modifications as an atomic patch. Each action is subject to two independent controls: the sandbox level (read-only or workspace-write) and the approval level (automatic or manual). These two settings are orthogonal and configured separately.

Codex CLI is aimed at developers comfortable with a terminal who want to delegate repetitive tasks, explore an unfamiliar codebase, or automate steps in CI pipelines, while keeping control over what actually runs.

It is not suited to people who prefer a graphical interface, or to those who do not want to manage sandbox and approval settings to secure agent actions.

What it offers

  • ·Shell command execution directly on the local machine from a Codex session.
  • ·Atomic file patching via the apply_patch command.
  • ·Sandbox level control: read-only or workspace-write mode, configurable independently.
  • ·Approval level control: automatic or manual validation of each agent action.
  • ·Interactive terminal mode or non-interactive mode via codex exec for pipelines and scripts.
  • ·Integrated code review: analysis of uncommitted changes, a commit, or a base branch without modifying the working tree.
  • ·Connection to plugins and skills to extend agent capabilities without leaving the CLI. (paid)

In practice

Price
Free
Commercial use
Commercial use: yes
Available on
codex
Host
codex
FR interface
No

Security and data

Rating
Security: serious reservations
Capabilities
readswritesexecutesnetwork
Your data
Data stays local
Maintenance
Active
Last release
23 October 2026

Links and repository checked on 25 October 2026 · added on 1 April 2025

How to install shell, apply_patch (Codex)

1. Open a terminal on macOS, Linux, or Windows. 2. Install Codex CLI using the appropriate install command for your system (see install_cmd for macOS/Linux). 3. Open a project directory and run the codex command. 4. On first launch, choose Sign in with ChatGPT to use your existing plan, or configure an API key if you prefer that option. 5. Before any write task, check the active sandbox level via /permissions to avoid unwanted file modifications. 6. Create a Git checkpoint before starting a task so you can revert changes if needed.
curl -fsSL https://chatgpt.com/codex/install.sh | sh

Official documentation ↗

In the same category

Whole category Code →