> ## Documentation Index > Fetch the complete documentation index at: https://docs.opper.ai/llms.txt > Use this file to discover all available pages before exploring further. # Control Plane > The five tools that make AI behavior in production observable, governable, and improvable. Once a model works, you usually need it to work reliably under your own policies. That's what the Control Plane is for. You attach a set of rules to your organization or a project, and every call that flows through Opper passes through them. ## The five tools Score every response against criteria you write. Pick how often to check and how strict to be. Pin a default model for an organization or project. Callers can still override. Block or redact content before it reaches the model and before responses go back to your app. Limit which models can run, how long traces stick around, and what you spend. Use Observe scores and feedback to pick better examples and tune prompts. ## How rules work Every Control Plane feature is a list of rules. Each rule has: * a **config**: the rule's settings (regex, allowlist, threshold, etc.) * an optional **name** * a **scope**: organization or project You add rules in **Controls → \** in the platform. They fire at runtime. Each one shows up on the call's trace, labeled with the feature, the rule's name, and the scope it came from. The playground has a **Project controls** toggle so you can test how a call would behave with or without your project's rules. ## How scopes layer Rules layer downward. An organization rule applies to every project; a project rule applies only to that project. For allowlists (Comply), scopes intersect. A model must be allowed at both levels to be reachable.