When initializing the MCP server we give the model a list of tools that it can use. After #273 / #274 the auth credential it has may make it so it can't use many of these tools and would be wasted context and likely will confuse the model if it tries to call them. Instead we can call the get_permissions endpoint on startup and then only expose those to the model.
When initializing the MCP server we give the model a list of tools that it can use. After #273 / #274 the auth credential it has may make it so it can't use many of these tools and would be wasted context and likely will confuse the model if it tries to call them. Instead we can call the
get_permissionsendpoint on startup and then only expose those to the model.