Codex cloud

You will like the handoff: comment on a GitHub pull request, a Linear issue or a Slack thread, and the task runs somewhere else, then comes back as a diff you can turn into a pull request. It is already inside the ChatGPT Plus plan, so the marginal cost of trying it is zero. The daily trait is asynchronous: you queue the small stuff before lunch and review it after, which changes how you triage a backlog.

Not for anything that needs your local environment or a database only your laptop can reach. Pick it for the backlog. Pick Jules if your stack is Google and your issues live there.

The risk is the environment. The docs have you configure dependencies, environment variables and secrets per repository and then set an internet access policy, so a task with the wrong policy is a process holding your credentials on an open network, running code a model wrote. The policy is set once, by whoever set up the repo, and forgotten by everyone who delegates a task afterward.

Use scoped tokens in the environment, never your own, and default the network to off. What it does right: every task gets a dedicated environment, so one bad run does not contaminate the next.

The pipeline is conventional and complete. 1. The repository is cloned into an environment configured with setup steps. 2. Many tasks run in parallel. 3. Output is a summary and a diff, with task logs as the verification surface, which means verification is whatever the agent chose to run and the reader chose to read. It began as a research preview on May 16, 2025 running codex-1, and no benchmark is published for the current models.

The consequence is that quality claims rest on the logs, task by task. The observation: the logs are the methodology, and most users will not open them.

The company is OpenAI, so strategy is the question, not survival. The Codex app merged into the ChatGPT desktop app in July 2026, which says the agent is a subscription feature, not a product line, and features get whatever roadmap the subscription needs. Pricing power is real: Pro starts at $100 with 5x or 20x usage tiers, and a ladder like that exists because people climb it.

No acquirer; the pivot already happened, from app to tab. Position: the safest closed bet on this board, with zero leverage for you, so negotiate nothing and plan for the tab to move again.

The demo is a task queue in the cloud. Procurement: Business at $20 per user is $1,200 a month for sixty, extra usage is bought as credits priced per model, and Enterprise and Edu tiers sit above, the shape procurement already knows from ChatGPT and the same admin console. The account is the ChatGPT workspace we already administer, so identity is solved by inheritance. Headless runs suit CI.

Onboarding is a repository connection and an environment file, an afternoon per repo. Approved with conditions: a credit ceiling per workspace before rollout, and secrets scoped per environment.

Nothing to own. One vendor's models, no key field, and the pricing doc states that API keys do not unlock cloud features, so bring-your-own is not a path. No local models, no source, and the environment runs in their cloud, so I cannot inspect the container it uses. There is an MCP client, documented under extend/mcp, the one place I can add a tool of mine.

That is the entire surface: a socket for tools, on a machine I cannot see, billed in a unit I cannot buy with a key. It works until it does not, and when it does not I wait.