Uploaded December 2025 | Updated September 2026, 6 hours ago
Model Context Protocol (MCP) is moving fast — from developer tools and coding agents to consumer apps and AI-driven workflows. But what works for experimentation doesn’t automatically work for the enterprise.
In this talk, Tobin South (WorkOS) breaks down what it actually takes to make MCP enterprise-ready — and why security, identity, and standards are the real unlock for adoption.
You’ll learn:
• Why MCP’s current developer-first model doesn’t map cleanly to enterprise risk tolerance
• How non-deterministic AI, dynamic tools, and agent identities change the security model
• Why OAuth 2.1 is becoming a baseline requirement for MCP servers and AI integrations
• What enterprises expect by default: SSO, SCIM, audit logs, event streams, and access controls
• How identity is expanding beyond humans to include AI agents — and why that matters
• Where MCP fits in the broader AI infrastructure stack as it moves toward production
This isn’t a product demo or a hype talk. It’s a practical look at how MCP evolves from “cool and useful” to something enterprises can actually deploy — safely, predictably, and at scale.
WorkOS exists to solve the enterprise side of this transition: authentication, identity, security, and governance — so teams can adopt new AI standards without reinventing the basics or taking unnecessary risk.
If you’re building MCP servers, integrating AI agents, or evaluating MCP for enterprise use, this talk outlines the questions you should already be asking — and the infrastructure you’ll need to answer them.
Learn how enterprises secure MCP with OAuth and identity at workos.com/mcp
Model Context Protocol (MCP) is moving fast — from developer tools and coding agents to consumer apps and AI-driven workflows. But what works for experimentation doesn’t automatically work for the enterprise.
In this talk, Tobin South (WorkOS) breaks down what it actually takes to make MCP enterprise-ready — and why security, identity, and standards are the real unlock for adoption.
You’ll learn:
• Why MCP’s current developer-first model doesn’t map cleanly to enterprise risk tolerance
• How non-deterministic AI, dynamic tools, and agent identities change the security model
• Why OAuth 2.1 is becoming a baseline requirement for MCP servers and AI integrations
• What enterprises expect by default: SSO, SCIM, audit logs, event streams, and access controls
• How identity is expanding beyond humans to include AI agents — and why that matters
• Where MCP fits in the broader AI infrastructure stack as it moves toward production
This isn’t a product demo or a hype talk. It’s a practical look at how MCP evolves from “cool and useful” to something enterprises can actually deploy — safely, predictably, and at scale.
WorkOS exists to solve the enterprise side of this transition: authentication, identity, security, and governance — so teams can adopt new AI standards without reinventing the basics or taking unnecessary risk.
If you’re building MCP servers, integrating AI agents, or evaluating MCP for enterprise use, this talk outlines the questions you should already be asking — and the infrastructure you’ll need to answer them.
Learn how enterprises secure MCP with OAuth and identity at workos.com/mcp
