Morgan Stanley is overhauling its API program to support AI agents by adopting CALM (Architecture as Code) and FINOS CALM, enabling secure and compliant deployment of APIs as stable, governance-ready building blocks behind evolving AI tools.

APIs are positioned as adaptable, secure, and governable interfaces that balance experimentation with reliability as AI tools and protocols evolve.

Operational rollouts now push hundreds of platform updates monthly across all API platforms, delivering rapid, zero-downtime deployment of changes like resource limits and security patches.

Deployment gates enforce architectural consistency, security controls, and configurations across development, staging, and production, providing early CI/CD feedback and governance.

Agent-to-Agent (A2A) and MCP protocols guide how AI agents interact with internal services, while also introducing challenges in tool orchestration, disambiguation, and token-based costs.

Over 110 CALM-deployed APIs are live in production, significantly cutting deployment time from years to weeks by integrating architecture, platform, and security safeguards.

Model Context Protocol (MCP) enables AI agents to connect to tools and data for discovery, invocation, and validation of natural language queries, exemplified by an MCP server querying a Trades API.