Add OpenTelemetry metrics to vMCP server

[JAORMX]

Summary

Instrument the vMCP server with OpenTelemetry metrics to enable observability and monitoring of virtual MCP server deployments.

Background

ToolHive already has a comprehensive telemetry package (pkg/telemetry/) that provides:

• Distributed tracing with W3C Trace Context propagation
• Prometheus metrics export
• OTLP support for remote collectors
• HTTP middleware for automatic instrumentation

The vMCP server currently lacks this instrumentation, making it difficult to monitor and debug production deployments.

Reusable Components

Component	Location	Description
Telemetry Middleware	pkg/telemetry/middleware.go	HTTP instrumentation
Config	pkg/telemetry/config.go	Configuration model
Providers	pkg/telemetry/providers/	OTEL provider setup

Existing Metrics (from pkg/telemetry/middleware.go)

toolhive_mcp_requests_total      // Total requests by method, status, MCP method, server, transport
toolhive_mcp_request_duration_seconds  // Request latency histogram
toolhive_mcp_active_connections  // Active connections gauge
toolhive_mcp_tool_calls_total    // Tool-specific call counts

Implementation

1. Add telemetry middleware to vMCP server (pkg/vmcp/server/server.go)

   • Wrap HTTP routes with telemetry.HTTPMiddleware
   • Configure service name as toolhive-vmcp

2. Add vMCP-specific metrics

   toolhive_vmcp_backend_requests_total     // Requests per backend
   toolhive_vmcp_backend_latency_seconds    // Per-backend latency
   toolhive_vmcp_composite_tool_duration_seconds  // Composite workflow execution time
   toolhive_vmcp_backends_discovered        // Gauge of active backends
   

3. Add configuration support

   • CRD: Add telemetry config to VirtualMCPServerSpec
   • CLI: Add telemetry flags to vmcp serve command

4. Expose /metrics endpoint

   • Prometheus scrape endpoint for metrics collection

Acceptance Criteria

• [ ] vMCP server uses existing telemetry middleware
• [ ] Prometheus metrics exported at /metrics endpoint
• [ ] Per-backend request metrics tracked
• [ ] Composite tool execution metrics tracked
• [ ] Telemetry configurable via CRD and CLI
• [ ] Sensitive data (tokens, secrets) redacted from traces
• [ ] Unit tests for metrics emission
• [ ] Documentation updated