So you built it with AI. It runs, it looks good, now you need to host it online and you'd prefer something Canadian.
What each tool actually hands you
Cursor and Claude Code are fully ready to plug into Canner. In fact, if you create a Canner account and get an API key, you don't even need to use the UI. Claude, Cursor, Grok, or whatever vibe-coding tool you're using can connect directly to our CLI and get your project online.
Bolt.new's exports are messy. Lovable ties everything to Supabase. Replit's exports run fine locally, but assume Replit's own env variables. But they can all get into Canner.
Getting the messy ones in
Create a project in the dashboard, and try dragging a zipped copy of the folder into the upload panel (or click Upload archive and select it). If you create a GitHub repo from the archive, you can also connect that repo and either run canner deploy from the CLI or let the project deploy automatically once connected. All three methods do the same thing.
What usually goes wrong
Hallucinated dependencies are the most common AI-coding problem. A package name that sounds real but was actually training data from years ago. Canner checks each one against the actual registry and discards the rest. Missing lockfiles are normal if you've never run it locally. Hardcoded API keys get flagged before anything ships, so you'll need to add those too. The important thing here? Don't panic. Copy the error messages and deliver them to your AI agent, or ask an agent. Sometimes Lovable and the others will actually help you package the app for Canner.
Sometimes they refuse - this is the American platform lock-in strategy, and a key reason you're probably trying to move away from them in the first place.