---
title: "intentic vs Codex · a harness intentic runs"
description: "OpenAI Codex is one of intentic's built-in harnesses. Its cloud runs on OpenAI's machines; intentic runs every harness in a sandbox on yours. Free and MIT."
url: "https://intentic.dev/compare/codex/"
---

[Comparisons](https://intentic.dev/compare/) Agent CLIs · intentic runs these

# Codex is one of intentic's built-in harnesses. Its cloud runs on OpenAI's machines; intentic's on yours.

Codex is both a CLI agent and a cloud coding platform inside ChatGPT. intentic runs its CLI as one of six harnesses, so this compares the cloud platform with the sandbox around it.

> "The same powerful coding agent — now in ChatGPT." A command centre for agentic coding with cloud environments, worktrees, scheduled tasks and an always-on background mode.

*Codex ·[openai.com](https://openai.com/codex)*

[Create your workspace](https://app.intentic.dev)[See the product](https://intentic.dev/features/)

The short answer

A strong agent embedded in ChatGPT, with cloud environments, parallel worktrees, skills you teach it, scheduled runs and a CLI that works offline. The IDE extension and desktop app round out a polished surface.

Codex is also one of intentic's built-in harnesses: run it beside Claude Code, Grok, Kimi Code and Google, inside a container you approve, with your own capabilities and automations.

## Head to head

1 of 12 rows go to Codex: ▸ marks them

Where you agree · both isolate parallel agents in git worktrees, both run on schedules and events, and both use the model subscription you already pay for. intentic runs Codex, so the agent can be the same one.

| Dimension | intentic | Codex |
| --- | --- | --- |
| What it is | the machine and fleet an agent works in | a coding agent inside ChatGPT with cloud environments |
| Where the agent runs | a Docker sandbox on your hardware | OpenAI's cloud, or locally via the CLI and IDE extension |
| Harnesses / models | Claude Code, Codex, Grok, Kimi Code, Google, plus any ACP agent | OpenAI models including GPT-6, on your ChatGPT plan |
| Surfaces | any browser over a private tunnel, phone included | ▸ChatGPT web and desktop, IDE extension, CLI |
| What you can change about the environment | the image, the capabilities, the context loaded each turn | skills, AGENTS.md, plugins; the cloud environment is OpenAI's |
| Where credentials live | inside a sandbox you own, injected per turn | with OpenAI for the session's duration |
| Starts on an event | automations: cron, webhook, chat, email, CI/CD, workspace events | scheduled tasks, GitHub Action, hooks, workspace agents |
| Licence | MIT for the sandbox, platform and CLI | proprietary; CLI is open source |
| Price | free, on your own model subscription | bundled with ChatGPT plans; Pro from $100/mo for full Codex |
| Parallel agents | a fleet of worktree-isolated agents across your sandbox | parallel agents in cloud worktrees and environments |
| Runs without you at the keyboard | automations on cron, webhook, push, alert, email, chat | scheduled tasks, always-on background mode |
| Review workflow | the diff in the Changes panel, approved before it lands | code review inside ChatGPT, PRs from cloud sessions |

## What it comes down to

01

### Whose machine the cloud runs on

Codex cloud environments run on OpenAI's infrastructure, and your repo is cloned into them. An intentic sandbox is a container on a machine you choose; the platform stores identity and a URL, never your code.

02

### One harness, or the workshop around it

Codex is the harness. intentic is the machine and fleet around one: six side by side, an image you approve, capabilities that carry your systems, a board that sorts your agents, and automations that wake them.

Not either/or

## Run Codex inside intentic

Pick Codex as an agent's harness in intentic and it runs inside a sandbox you own, with your capabilities and automations around it. The next agent can use Claude Code or Grok without switching tools.

## Pick Codex when

You want one subscription that covers the model, the cloud environment and the desktop app, you prefer OpenAI's models, and you would rather not run your own container.

## How this was checked

Every claim about Codex on this page is drawn from its own documentation, last read on 18 August 2026. These products move fast. Spot something out of date? [Open an issue](https://github.com/intentic/intentic/issues) and it is fixed in the next build.

- [Codex · openai.com](https://openai.com/codex)
- [Codex · openai.com](https://openai.com/codex)
- [CLI source · github.com](https://github.com/openai/codex)

[Next comparison vs Conductor → Parallel agents on a Mac versus a configurable sandbox for the fleet](https://intentic.dev/compare/conductor/)

[Create your workspace](https://app.intentic.dev)[All comparisons](https://intentic.dev/compare/)