Tool filtering is the practice of limiting which tools an AI agent can discover or use based on the agent’s identity, the user it represents, applicable policy, and the circumstances of the request. In Model Context Protocol environments, filtering may affect the tool definitions returned during discovery, the tool calls permitted at runtime, or both. Its purpose is to reduce unnecessary agent capabilities and prevent access to tools that fall outside the agent’s assigned role or current task. Because hiding a tool does not by itself prevent its invocation, tool filtering should be paired with authorization and enforcement on every tool call.
