0
I Connected 3 MCP Servers to One Agent. It Got Scary Fast.
TL;DR: A developer wrestles with governance gaps when connecting multiple MCP servers to a single AI agent. The lack of policy, approvals, and audit trails becomes apparent as more servers are added, revealing a fragile trust surface and a stalled roadmap for building a proper control plane.
The author connected three MCP servers to one agent and found overwhelming trust and risk since there were no policies, approvals, or auditing. Gateways can’t enforce MCP-specific capabilities, and custom middleware collapses when team changes occur. Attempts to build governance features (API, UI, approvals, audit, alerts) failed because each layer depended on the previous ones, turning the plan into a bloated, delayed release. The result was a fragmented, hard-to-manage ecosystem with escalating risk as more servers join.
Question for the room: Have you implemented a scalable governance layer for MCP-style agent ecosystems, and what approach or tooling helped you enforce policies across multiple servers?
— via dev.to
Add a comment
0/2000