What happened
On Oct 8, 2026 Okta announced that Agent Gateway, the runtime enforcement layer within Okta for AI Agents, is now generally available. It validates every agent-to-tool (MCP) call against identity policies using RFC 8693 token exchange and will not process a request unless the credential names BOTH the acting agent and the human user, preventing confused-deputy attacks and unauthorized permission inheritance. It also isolates credentials at the moment of action and produces a unified audit trail.
Why it matters
This is one of the first major identity-vendor GA enforcement points that treats the agent-plus-user pair as the unit of authority — directly addressing the confused-deputy gap that plagues existing MCP gateways. For enterprises with agents on Claude Code/Copilot/Codex, it converts agent governance from logging to real runtime denial.
Applicability
Enterprises using Okta for AI Agents that connect agents to MCP servers and SaaS tools should enable Agent Gateway for agent-to-tool runtime policy enforcement now that it is GA.