OpenAI Codexcodex-cli 0.154 · macOS, Linux, WSL, Windows
1
Curious builder0 XP earned · 300 to level 2
0 daysFinish a lesson to begin
Badge collection0 of 6 unlocked
26 small wins to finish your pathNext lesson

Codex cloud

Everything so far ran on your machine. The cloud runs the same agent in an isolated environment, against your repository, while your laptop does something else.

What it gives youWhy it matters
Work in parallelLong tasks get their own environment and keep going
A reproducible environmentDependencies, tools, variables and setup steps, per repository
A review before you mergeA summary and a diff, a follow-up, or a pull request

The third row is the important one. Cloud work does not land in your repository by itself: you read the diff and decide, which is the same shape as the review in lesson 11.

Where a task can start

From the web, and from GitHub, GitLab, Linear or Slack. That is what makes it a different tool rather than a remote terminal: the work starts where the work is discussed, which is lesson 20.

The environment is the setup

A cloud task begins in an environment you configure: the dependencies it needs, the tools, the variables and the setup steps. Getting that right is most of making cloud runs reliable, and it is the same information your README should already contain.

  • Pin what you install. A task that installs a different version than you do produces different results.
  • Set the variables it genuinely needs, and nothing more.
  • Give it the test command. A task that can run your tests can check itself.
Internet access is a choice
Network access in cloud environments is its own setting, with its own documentation page. Decide it deliberately: a task that can reach the internet can also reach somewhere you did not intend.
Try it yourself
  • Set up one repository in Codex cloud and give it a small, well-described task.
  • Read the diff it produces rather than the summary, at least the first few times.
PreviousSubagents

This is what real progress feels like.