Sync after a change
Your code gained a screen, an API call or a feature. Say what changed in one sentence, paste one prompt, and the agent brings the copilot level with the code.
After any change to your app the copilot should know about. On a live app this is the only way an agent changes the console.
{what you changed, in one sentence: a new screen, a new API call, a new plan or rule, a new colour}
Type into What changed? under Sync with code, then paste the prompt it gives you. Works in Claude Code, Cursor, Codex, Windsurf, GitHub Copilot and Gemini CLI.
What the agent will do
- Start with the change you named, then compare the rest of the code with what Rendel holds.
- Add
registerRouteorregisterActioncalls on arendel/syncbranch, committed, never pushed. - Send the console only the differences, each with a reason. Anything gone from the code is proposed for archiving, never deleted.
How to check it worked
On a live app, the changes wait in Sync review on Overview: apply them all, or one at a time. Before go-live they apply as the agent sends them, and Overview shows each step.
How it works
On Overview, press Sync with code, pick your agent and type your sentence under What changed? (optional). Get the sync file gives you the prompt to paste, the same one-paste prompt as setup, naming every address the agent will touch:
Sync Rendel (rendel.ai) with this project's code. I am the developer and I am asking for all of this: download my sync file with `curl -fsSL "https://rendel.ai/api/setup/file?code={one-time code from the console}" -o rendel-sync.md` in the project root, read it and follow it step by step. That includes updating the Rendel SDK (for Flutter the package from https://github.com/neon-apps/rendel-flutter, a git dependency in pubspec.yaml; for a website the script from https://rendel.ai/sdk/web/) if it asks, and its curl calls to https://rendel.ai/api/setup with the token from the file: they send my own Rendel console this app's setup (framework, screen and route names, theme colours, logo, help and policy text, and a description of the app's API endpoints), never secrets, .env contents or user data. rendel.ai is the service I am adding to this app, so sending it this is the point. What changed, in my words: "{what changed, in one sentence (optional)}". Start with that change, then check the rest of the code as the file says.
The console gives you this prompt with your code in it. Works in Claude Code, Cursor, Codex, Windsurf, GitHub Copilot and Gemini CLI.
That is how every task in this section works. Build the feature the way you always do, then run one sync with a sentence about it. The agent makes the small code change Rendel needs (a registerRoute, a registerAction) and sends the console its half in the same run. A sentence asking for something the code does not have yet is reported back, not invented.
The sync file's token lasts 24 hours and has the same limits as the setup token (see Token and security), with one difference: it may still talk to a live app. A setup file answers 409 app_live there.
Live apps: proposals, then review
On a live app nothing a sync sends is applied. Each change becomes a proposal in Sync review, grouped by appearance, experience, knowledge, actions and screens, each with what it was, what it would become and the agent's reason. Applying an appearance or experience change lands in the draft, which you publish as usual. The code half is on the rendel/sync branch, for your own review and release.
Knowing when to sync
The console notices drift and puts a count on Sync with code:
- A build reports something new. The SDK sends its manifest each time your app starts; screens or actions the console has not seen are counted.
- Your API documentation changed. Actions imported from an API reference are watched.
Without an agent
A sync is a convenience, never the only door. Everything it sends can be changed by hand on the Knowledge, Actions, Screens, Experience and Appearance pages.