An MCP gateway is a centralized control plane between AI agents and MCP servers. It addresses prompt injection, tool poisoning, permission escalation, and inconsistent team configurations by enforcing authentication, applying access policies, logging requests, and blocking suspicious activity in real time. section 3.3.5
Use a gateway to make MCP access deliberate and auditable:
- Enforce authentication and access policies before requests reach MCP servers. section 3.3.5
- Log every request for audit trails and block suspicious activity in real time. section 3.3.5
- Pair centralized governance with a team-approved server allowlist and documented configuration. section 3.3.5
Figure 3.5 MCP Governance layers - the agent communicates through an MCP Gateway (enforcing auth and policies), which routes to an MCP Proxy (aggregating multiple servers into a curated toolset), which connects to individual MCP servers registered in a central catalog. Each layer adds security and reduces context noise.

The governance-layer diagram shows the agent communicating through a gateway, then a proxy and curated MCP servers, separating policy enforcement from tool aggregation. section 3.3.5

