OpenTelemetry MCP Server Configuration

The TalkOps OpenTelemetry MCP Server is configured using standard environment variables. You can pass these via .env files for local development, inject them through Docker flags, or mount them from Kubernetes ConfigMap and Secret objects in production.

---

Server & Transport Settings

Configure how MCP clients (Claude Desktop, Cursor, Cline, or backend orchestrators) connect to the server.

Variable	Default	Allowed Values	Description
MCP_TRANSPORT	stdio	stdio, http, sse, streamable-http	Communication protocol for MCP JSON-RPC messages.
MCP_HOST	0.0.0.0	Valid IP Address	Network interface to bind for HTTP/SSE transports.
MCP_PORT	8771	Valid TCP Port	Listening port for HTTP/SSE servers.
MCP_PATH	/mcp	URL Path	Base endpoint path for JSON-RPC message handling.
MCP_LOG_LEVEL	INFO	DEBUG, INFO, WARNING, ERROR	Logging verbosity for operational diagnostics.
MCP_LOG_FORMAT	json	json, text	Structured JSON or human-readable plain text output.
MCP_HTTP_TIMEOUT	300	Integer (Seconds)	Maximum client request duration before timing out.

---

Kubernetes & Operator Integration

The server directly queries and modifies OpenTelemetry Operator custom resources. Adjust these if your cluster uses customized CRD group names or legacy API versions.

Variable	Default	Description
K8S_ENABLED	true	Enable or disable Kubernetes API interactions. (Set false for standalone registry testing).
K8S_IN_CLUSTER	false	Set to true when running inside a Pod with an attached ServiceAccountToken.
OTEL_CRD_GROUP	opentelemetry.io	API Group name for OpenTelemetry Custom Resource Definitions.
OTEL_CRD_API_VERSION	v1beta1	API version for OpenTelemetryCollector CRDs.
OTEL_INSTRUMENTATION_API_VERSION	v1alpha1	API version for Instrumentation CRDs (promoted independently from Collectors).
OTEL_COLLECTOR_PLURAL	opentelemetrycollectors	Plural resource name for Collector CRDs.
OTEL_INSTRUMENTATION_PLURAL	instrumentations	Plural resource name for Instrumentation CRDs.
OTEL_TA_SERVICE_DISCOVERY	true	Enable Target Allocator service discovery across namespaces.
OTEL_TA_DEFAULT_PORT	8080	Default listening port for Target Allocator instances.

---

Backend Discovery & Language Registry

Variable	Default	Description
PROMETHEUS_BASE_URL	(empty)	Optional HTTP endpoint for Prometheus (used to validate live series counts during cardinality audits).
PROMETHEUS_TIMEOUT	30	Request timeout for Prometheus queries in seconds.
PROMETHEUS_VERIFY_SSL	true	Enable SSL certificate verification for remote backend queries.
OTEL_LANG_REGISTRY_PATH	(builtin)	Custom filesystem path to override the bundled language support matrix JSON.

---

Kubernetes RBAC Requirements

When running in-cluster or using intent-driven collector provisioning (otel_provision_collector), the MCP server's ServiceAccount needs permissions to manage OTel CRDs, inspect Pods/Deployments, and create the necessary ClusterRole / ClusterRoleBinding resources for the k8sattributes processor:

apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRole
metadata:
  name: talkops-opentelemetry-mcp-role
rules:
  # OpenTelemetry Operator CRDs
  - apiGroups: ["opentelemetry.io"]
    resources: ["opentelemetrycollectors", "instrumentations"]
    verbs: ["get", "list", "watch", "create", "update", "patch", "delete"]
  
  # Workload inspection & auto-instrumentation injection
  - apiGroups: ["apps"]
    resources: ["deployments", "daemonsets", "statefulsets"]
    verbs: ["get", "list", "watch", "update", "patch"]
  
  # Pod, namespace, and service discovery
  - apiGroups: [""]
    resources: ["pods", "services", "namespaces", "configmaps", "secrets"]
    verbs: ["get", "list", "watch"]

  # Dynamic RBAC creation for k8sattributes enrichment
  - apiGroups: ["rbac.authorization.k8s.io"]
    resources: ["clusterroles", "clusterrolebindings"]
    verbs: ["get", "create", "patch"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: ClusterRoleBinding
metadata:
  name: talkops-opentelemetry-mcp-binding
subjects:
  - kind: ServiceAccount
    name: talkops-opentelemetry-mcp-sa
    namespace: monitoring
roleRef:
  kind: ClusterRole
  name: talkops-opentelemetry-mcp-role
  apiGroup: rbac.authorization.k8s.io

---

Production Deployment Manifest

Here is a ready-to-apply Kubernetes Deployment manifest for running the OpenTelemetry MCP server in your monitoring namespace:

apiVersion: apps/v1
kind: Deployment
metadata:
  name: opentelemetry-mcp-server
  namespace: monitoring
  labels:
    app.kubernetes.io/name: opentelemetry-mcp-server
    app.kubernetes.io/part-of: talkops
spec:
  replicas: 1
  selector:
    matchLabels:
      app: opentelemetry-mcp-server
  template:
    metadata:
      labels:
        app: opentelemetry-mcp-server
    spec:
      serviceAccountName: talkops-opentelemetry-mcp-sa
      containers:
        - name: server
          image: talkopsai/opentelemetry-mcp-server:latest
          imagePullPolicy: IfNotPresent
          env:
            - name: MCP_TRANSPORT
              value: "http"
            - name: MCP_PORT
              value: "8771"
            - name: K8S_IN_CLUSTER
              value: "true"
            - name: MCP_LOG_LEVEL
              value: "INFO"
          ports:
            - containerPort: 8771
              name: http-mcp
          resources:
            requests:
              cpu: 100m
              memory: 128Mi
            limits:
              cpu: 500m
              memory: 512Mi
          livenessProbe:
            httpGet:
              path: /mcp
              port: http-mcp
            initialDelaySeconds: 10
            periodSeconds: 15