Codex CLI

Codex CLI — OpenAI's open-source Rust agent, with the strongest local sandboxing on this page. Codex CLI is an Apache-2.0 Rust binary that runs agentic coding locally, with OS-level sandboxing via Seatbelt on macOS and Landlock on Linux — the most serious containment story of any local harness here. It pairs with Codex cloud tasks, which run in OpenAI's own containers and open pull requests. Although tuned for the GPT-5 Codex family, the config accepts arbitrary OpenAI-compatible providers, so local models work. Built by OpenAI. Founded 2025. Open source under Apache-2.0.

We track it in 1 comparison — Coding agents — so every claim below is a cell in a table you can open and check rather than an impression. Across those rosters it sits against 16 other tools, and what follows is where it visibly separates from them.

Where it wins.

• Opens PRs — Yes, which only 6 of 16 tools here manage.
• Background agents — Yes, which only 5 of 16 tools here manage.

Each of those is ranked against the whole roster on its category page, not against a hand-picked subset, so a first place here means first of everything we list.

Where it gives ground.

• Subagents — No. Claude Code records Yes.

None of these disqualify it on their own. They are the fields to check against your own requirements before you commit, because they are the ones where a competitor genuinely does better.

Provenance. 24 of 26 tracked fields carry a value for Codex CLI, and 8 of those cite a document you can open. Last verified 2026-01-15. Every figure keeps its own provenance — measured by us, claimed by the vendor, inferred, or community-reported — and we would rather print a dash than a guess.

Its nearest neighbour in our data is Aider. Codex CLI is ahead on Opens PRs (Yes against No) and MCP (Yes against No). Aider takes Model breadth (any provider / router against one vendor + custom endpoint) and Checkpoint & undo (Yes against Partial). That pattern repeats across the rest of the roster — see Codex CLI alternatives for the other rivals, each compared the same way.

Company

OpenAI

Founded

2025

Licence

Apache-2.0 (open source)

Fields we track

24 of 26

Last verified

2026-01-15

The best containment story of any local harness — Seatbelt and Landlock are real boundaries, not an allowlist with good marketing. Weaker than Claude Code on orchestration primitives: no subagents, notification-only hooks, thinner custom-command support.

Opens PRs

●Yes1st of 16

Inferredverified 2026-01-15

Subagents

○No10th of 14

Community-reported

Background agents

●Yes1st of 16

Inferredverified 2026-01-15

Entry price

$20 /seat/mo13th of 16

Vendor-claimedverified 2026-01-15source

Permissions

allowlist + OS sandbox1st of 16

Vendor-claimedverified 2026-01-15source

BYOK

●Yes1st of 16

Inferredverified 2026-01-15

Headless / CI

●Yes1st of 16

Vendor-claimedverified 2026-01-15source

MCP

●Yes1st of 15

Vendor-claimedverified 2026-01-15source

Codex CLI is the right call in these situations, each one drawn from a field we actually record:

• Unattended local runs where a real OS sandbox is non-negotiable.
• Teams already on ChatGPT business plans wanting agent use covered.
• Anyone who wants an Apache-2.0 harness they can audit and fork.
• Opens PRs is your binding constraint. Codex CLI records Yes, which only 6 of the 16 tools in the Coding agents roster do. We define that field as a first-class path to a pull request: a built-in command, a vendor-published GitHub App with its own identity, or a hosted runner.
• Background agents is your binding constraint. Codex CLI records Yes, which only 5 of the 16 tools in the Coding agents roster do. We define that field as can you fire off a task that keeps running once the client is closed — on vendor infrastructure or a detached local runner — and check back later?

Reach for something else when any of the following is a requirement rather than a nice-to-have:

• Subagents. Codex CLI records No on the Coding agents table. Claude Code records Yes on the same field. If that is a hard requirement rather than a preference, start elsewhere.

We publish this block because a comparison that only lists what a tool is good at is marketing. Every figure above sits on the same page as its source, and the field definitions are on the category tables if you want to check how we measured them.

What we don't know. 2 of the 26 fields we track for Codex CLI are still blank: Terminal-Bench and Bench model. Those render as dashes rather than as zeroes or assumptions, because an empty cell and a bad cell are not the same thing and only one of them is honest. If you know any of these figures and can point at a document, tell us.