When using multiple AI Agents that need to call the same external APIs (e.g., order lookup, CRM sync, appointment booking), each agent requires its own copy of the HTTP request action configured independently. This means any change to an API endpoint, headers, or authentication must be updated manually in every agent that uses it. This creates maintenance overhead and increases the risk of inconsistencies as the number of agents grows.

A workspace-level library of HTTP request actions that can be created and maintained in one place, then selectively enabled or disabled per AI Agent — similar to how Knowledge Sources work today. Each agent would toggle on only the HTTP actions it needs, while the actual API configuration (URL, method, headers, body, input fields) is managed centrally.