OpenAI

Codex meets StackShift.

Take your repository from local changes to a verified deployment.

1. Install the tools

You’ll need Codex installed with its model provider configured, plus the StackShift CLI. These are separate tools.

Already installed? Run stackshift setup agent --help to check that your CLI supports this command. Update the CLI if setup is not recognized.

2. Add guidance to your project

Open a terminal in your project’s root folder. Run this command in the terminal before starting a new agent conversation.

Terminal · setup command

stackshift setup agent --client codex

This installs StackShift’s skill in .agents/skills/stackshift-cloud-operations. It does not install Codex, connect your account or deploy anything. Existing custom guidance is preserved.

3. Sign in to StackShift

Run this in the same terminal and complete the browser sign-in. If you’re already signed in, use stackshift auth status to check your account.

Terminal · account sign-in

stackshift auth login

4. Open Codex

Open Codex in your project, or run codex in its terminal.

Terminal · open agent

codex

Already had it open? Start a new conversation in this project so it can discover the installed skill.

5. Give your agent a task

Paste this into the agent’s chat, not the terminal. It will inspect your project before making deployment changes.

Agent chat · first task

Use the StackShift skill to inspect this project and prepare it for deployment. Explain what needs to change. Do not deploy yet.

Existing application? Your working providers are preserved. Ask for deployment when you’re ready; data migration and production cutover are separate decisions.

Alternative: let your agent guide setup

If Codex is already open in your project, paste this into its chat. It will check the CLI and guide skill installation and sign-in. Start a new conversation afterward if needed to discover the skill.

Setup prompt

Read https://stackshift.cloud/prompt.md and use it to prepare this repository for StackShift. Inspect the project first, explain what it needs, and wait for my deployment request.

What will you build next?

Give your agent a concrete task. Start with one of these prompts and make it yours.

Prepare a project

Inspect this repository for StackShift deployment. Check its build and start commands, environment variable names, port and persistent data. Report what needs to change before deploying.

Deploy and verify

Deploy this project to my selected StackShift environment. Confirm the target and any new costs first, follow the build through readiness, and verify the live URL.

Investigate a failure

Inspect the latest failed deployment for this project. Find the first actionable error, explain its cause, and propose the smallest fix.

Connect a database

Check which database this app needs and whether my StackShift project already has one. Explain the configuration and cost before creating anything. Keep credentials out of chat.

Connect through MCP

Already have a connection? Keep using it. MCP gives your agent scoped tools; the skill adds project guidance.

MCP connection guides