I just run the agent directly on separate testing/dev servers via remote-ssh in VS Code to have an IDE to sanity check stuff. Just far simpler than local dev and other nonsense.