What started as a learning project to dive deeper into AI has turned into a product I feel will resonate with others administering GitHub at enterprise organizations.

https://gh-admin.com/ allows you to administer GitHub with natural language.

Simple things, like creating a new GitHub repository, turn into a chore when you need to assign teams, setup rulesets, etc. Often, copying settings from existing repositories and hoping you don't miss anything.

Now, what used to take more than a dozen clicks in the UI can be done with a single prompt, such as "Create the repos {org}/{repo1} and {org}/{repo2} from the {org}/{template_repo} template. The new repos should be {public|private}. After creating the repos, copy team access and branch protection from the template repo to the newly created repos."

For me, this is where MCP can be so much more than a simple API wrapper/proxy.

Built-in tools group multiple API calls into larger tasks that AI can understand, such as copying access or rulesets from one repo to another.

What are you all building with MCP?