The MCP Mesh
DevOps Engineer @ Accuris
Pulled five separate MCP servers (New Relic, ArgoCD, Snyk, Azure DevOps, and Scalr) into one interface engineers actually wanted to use.
Impact
- One interface for observability, deployments, security scanning, and infrastructure.
- Better query accuracy and lower latency than the old New Relic integration.
- Lower token usage on the custom New Relic MCP server.
The problem
Our engineering work was spread across five separate tools, each with its own MCP server and its own quirks. Jumping between New Relic for observability, ArgoCD for deployments, Snyk for security, Azure DevOps for tickets, and Scalr for infrastructure added up to a real daily tax.
The approach
I built an MCP mesh that put all five behind one interface. Instead of five disconnected endpoints, engineers got one place that routed each request to the right backend and handed back a consistent response, so cross-tool workflows finally worked.
While I was at it, I wrote a custom New Relic MCP server that beat the one we were using. It generated more accurate queries, used fewer tokens, and came back faster.
The impact
- One interface for observability, deployments, security scanning, and infrastructure.
- Noticeably better query accuracy and latency than the old setup.
- Lower token cost per request, which lowered what the tooling cost to run.