CodeYam Editor in GitHub Codespaces (staging)

This is the staging starter. It installs the editor's staging npm dist-tag, which carries changes that have not been released yet, so you can try them in a real codespace before they ship. For the released editor, use codeyam-ai/codespaces-starter. Apart from that channel, the two templates are identical.

Click the button to try the CodeYam Editor with nothing installed on your own machine. It runs in a GitHub codespace, a Linux machine GitHub hosts for you and opens in your browser.

What happens when you click it

GitHub asks for nothing when it creates the codespace: there is no API key to fill in. You sign in to the AI inside the editor.

1. The codespace installs everything. That means the editor, the Claude Code CLI, and a browser for screenshots. The first time takes a few minutes. A welcome page explains what is happening while you wait.
2. Wait for the link in the terminal. When the editor is ready, the terminal panel at the bottom of the window shows a boxed "CodeYam Editor is ready" with a link. Cmd+click it (Ctrl+click on Windows or Linux).
3. The editor opens on the New Project screen. Pick the AI that should build, sign in to it, then describe what you want to build. The editor builds it with you against a live preview.

This repository is intentionally almost empty: no app, no .codeyam/. Your project is created inside the codespace the first time the editor starts.

Keeping your work

The button makes a new codespace from this template, not one on it, so your project starts with no repository of its own and this template is never changed. After your first build, the editor offers to create the project's own GitHub repository (private by default, under your account). It asks you to sign in to GitHub once, inside the editor, because the codespace's built-in token cannot create repositories. The new repository's README gets its own Open in GitHub Codespaces button, which reopens that project.

Good to know

• Who can see the editor: only you. GitHub keeps the editor's port private and asks you to sign in to GitHub before it lets anyone through. If you make the port public, the editor starts requiring its own one-time token.
• Stopping and restarting: your files are kept, and the editor comes back on the same address when the codespace restarts.
• Signing in to Claude: the editor shows a link. Open it, approve access, and paste the code it gives you back into the editor. A rebuilt codespace asks you to sign in again; a stopped and restarted one does not.
• Using Codex, Gemini or OpenCode instead: pick it in the editor and paste its API key there. The key is saved with your project. If you already keep it as a Codespaces secret (OPENAI_API_KEY, GEMINI_API_KEY or OPENCODE_API_KEY under GitHub's Settings → Codespaces → Secrets, with access to this codespace's repository), the editor uses it and skips that step.

How this template is made

The files in .devcontainer/ (devcontainer.json, the welcome page codeyam-welcome.md, and codeyam-ready.sh, which prints the editor's link) are generated by codeyam-editor editor devcontainer-setup --channel staging run in an empty directory. They are not written by hand, so they match what existing projects get. The editor repository runs a drift check against them (CODESPACES_STARTER_CHANNEL=staging scripts/check-codespaces-starter-drift.sh). To regenerate them, run codeyam-editor editor starter-refresh --apply --channel staging in a checkout of this repository and commit .devcontainer/. When you change this README, make the same change in codeyam-ai/codespaces-starter unless it is about the channel.