Multicorn Shield – Open-source permissions and approvals for AI agents

Multicorn Shield is an open-source permissions layer for AI agents that addresses security risks when agents access email, calendars, bank accounts, and code repositories. The tool provides an SDK to control agent actions through consent screens, scoped permissions, and activity logging. It offers three integration options: wrapping existing agents as a proxy, native OpenClaw integration, or direct SDK implementation. Shield intercepts tool calls, requires user authorization for new actions, logs all activity to a dashboard, and detects destructive commands requiring separate approval.