|
Flipper Cloud now runs a Model Context Protocol (MCP) server. Connect your AI assistant and ask it "Which Flipper flags can we remove?" Most feature flags are a promise to come back and clean up later. Most teams never do, and old flags pile up until nobody remembers which ones are safe to delete. Your assistant can now keep that promise for you. How does it decide a flag is done?It looks at every environment in a project together, because a flag isn't done until it's settled everywhere. Each candidate comes with its reason and how confident to be:
Flags changed in the last week and flags marked permanent are left alone. Then your assistant can do the tedious part: remove the check from your code, keep the right code path, and open the PR. How do I connect?Add
Cursor, Codex and anything else that speaks MCP work too, with sign-in or an API key. The MCP docs have the setup for each. What can it see?Only what you can see in Flipper Cloud. Your teammates' email addresses and your customers' actor IDs never leave Flipper Cloud. You can disconnect an assistant anytime under Profile → API Keys. What's next?For now the server is read-only. Next up are writes: ask your assistant to add, enable, disable or remove a flag, with your project role still deciding what it's allowed to do. That's coming soon. Give it a try and let us know which flags it finds. Hope you find this helpful and keep the feature requests coming. |