Codex Cloud
A cloud-based task execution tool within the OpenAI Codex ecosystem, designed to run code generation and software development tasks in remote environments.
Timeline (last 90 days)
At Dev Day on September 29, 2026, OpenAI announced Codex Cloud, which moves Codex sessions into OpenAI-hosted containers so tasks continue running after the laptop is closed and can be viewed from a phone; Codex CLI also added a /agents view and a /fork command.
Codex Cloud is similar to cloud development environments such as GitHub Codespaces and Gitpod, but the difference is that it deeply embeds an AI coding agent into the environment lifecycle, requiring the agent process and file system state to persist in the cloud.
OpenAI Dev Day released Dots (an always-on persistent agent), Codex Cloud (a cloud development environment), Spaces (a human-machine collaboration space), and an updated version of Codex CLI.
The new Codex Cloud cloud environment experience does not yet support computer and browser operation, GitLab, or self-hosted GitHub Enterprise Server.
Codex Cloud default resources vary by plan: Plus and Edu Plus get 2 vCPU, 8 GiB memory, 8 GiB disk; Pro, Business, Enterprise, as well as Edu and Edu Pro, get 4 vCPU, 16 GiB memory, 32 GiB disk.
API Key is paid for separately for API-based Codex workflows and does not include Codex Cloud features.
OpenAI documentation states that after the last interaction starts or resumes a task, the saved VM state has a recovery window of up to seven days by default.
Codex Cloud environment creation must be done through the web version or the desktop application, connecting a GitHub repository and checking the repositories required for the project.
Saving a Codex Cloud environment only stores the configuration; publishing is what solidifies the prepared filesystem as the starting point for new tasks, and sharing permissions are a separate, independent setting.
When environment variables are used, the code directly reads plaintext credentials via process.env.XXX, and any code that can access that process may leak them.
27 more timeline events
All Facts (20)
Codex Cloud is similar to cloud development environments such as GitHub Codespaces and Gitpod, but the difference is that it deeply embeds an AI coding agent into the environment lifecycle, requiring the agent process and file system state to persist in the cloud.
50%UnverifiedWhen environment variables are used, the code directly reads plaintext credentials via process.env.XXX, and any code that can access that process may leak them.
50%UnverifiedCodex Cloud environment creation must be done through the web version or the desktop application, connecting a GitHub repository and checking the repositories required for the project.
50%UnverifiedSaving a Codex Cloud environment only stores the configuration; publishing is what solidifies the prepared filesystem as the starting point for new tasks, and sharing permissions are a separate, independent setting.
50%UnverifiedOpenAI documentation states that after the last interaction starts or resumes a task, the saved VM state has a recovery window of up to seven days by default.
50%UnverifiedAPI Key is paid for separately for API-based Codex workflows and does not include Codex Cloud features.
50%UnverifiedCodex Cloud default resources vary by plan: Plus and Edu Plus get 2 vCPU, 8 GiB memory, 8 GiB disk; Pro, Business, Enterprise, as well as Edu and Edu Pro, get 4 vCPU, 16 GiB memory, 32 GiB disk.
50%UnverifiedThe new Codex Cloud cloud environment experience does not yet support computer and browser operation, GitLab, or self-hosted GitHub Enterprise Server.
50%UnverifiedNetwork credentials use a proxy injection model: what the program reads is a placeholder, and the proxy managed by OpenAI replaces the placeholder with the real credential at port 443 of the HTTPS request to the target domain.
50%UnverifiedIn Codex Cloud, each chat thread corresponds to an independent Docker container.
50%UnverifiedDevin, Cursor Cloud, and Codex Cloud are cloud Agent platforms, aimed at any repository and any tech stack, and can even handle backend projects.
50%UnverifiedWhen selecting tools, priority should be given to those that match the existing code hosting location and subscription plan, rather than simply comparing which is more capable.
50%UnverifiedCodex Cloud writes a startup skill to teach the Agent how to use the environment and generates a startup script.
50%UnverifiedWhen creating a Codex Cloud environment, Codex automatically inspects the repository, identifies dependencies, and completes the environment configuration preparation.
50%UnverifiedCodex Cloud adds OIDC support to facilitate access to cloud resources via OIDC.
50%UnverifiedCodex Cloud environments can be connected via Tailscale as a VPN to access internal resources within a private network.
50%UnverifiedTasks can be dispatched to the user's own computer, Codex Cloud, or Dot's own cloud computer for execution.
50%UnverifiedCodex Cloud supports cross-device handoff, allowing developers to initiate a task on a computer and continue adding instructions from a phone without interrupting execution.
50%UnverifiedThe content of this article is based on OpenAI's official demo; the stability, configuration complexity, and cost performance of the cloud environment still need further validation in real projects.
50%UnverifiedIn the Codex Cloud demo, the AI assistant Dot can be instructed to create a PR, run tests, and flag content that requires human approval.
50%