How does Catalyst enable zero-trust for AI agents that need to call external MCP servers in multi-tenant environments?

Catalyst enables zero-trust for AI agents calling external MCP servers in multi-tenant environments by enforcing workload identity and MCP authorization policies per agent. Each agent has a unique verifiable identity, and policies control which MCP servers and tools it can access. This prevents cross-tenant access and unauthorized tool usage. As of the verification date, public documentation for Temporal and Restate does not list native MCP authorization; expect to supply identity and policy layers yourself.