Should AI tools write feature flags for you?

We launched Bucket MCP last week and got featured in The Leaderboard.

@aaronoleary put it elegantly:

Whether you're using @Cursor, @Augment Code, or @Windsurf, you can just tell it what you want flagged, and the MCP server handles it. So instead of context-switching to write config files or CLI commands, you just prompt: “flag the download button.” That’s it.

I'd love to follow up here and hear from the community — Is it too much trust to hand over?